Statios

Capability Statement

Modern statistical analysis — desktop & web, AI-assisted, reproducible.

Generated from the engine handler inventory · 600 procedures

Statios unifies the analytical scope of SPSS, Stata and EViews in one modern application, with a built-in AI assistant and one-click, citation-ready reporting. Every figure is computed by an open, auditable Python engine that runs both on the desktop and entirely in the browser.

600
statistical procedures
11
analysis domains
569
fully supported
31
with documented limits

What we do

DomainProceduresScope
Descriptive Statistics & Hypothesis Tests37Descriptive statistics and statistical hypothesis tests covering parametric (t-tests, ANOVA), non-parametric (Mann-Whitney, Kruskal-Wallis, Friedman), correlation, and categorical association tests.
Regression & Generalized Linear Models142Statios's model namespace covers 50+ regression and generalized linear estimation procedures, from OLS to latent-variable extensions. Methods range from standard parametric models (linear, logistic, count) to specialized estimators for endogenous/selected samples, censoring, and complex errors. All predictor-taking handlers auto-expand categorical predictors R-style (labeled indicator columns, named reference level, factors field) — no manual dummy coding required.
Post-Estimation & Diagnostics47Statios provides 31 post-estimation and regression diagnostic procedures spanning heteroskedasticity tests, serial correlation diagnostics, influence measures, marginal effects, contrasts, and model comparison tools.
Time Series & Macroeconometrics84Comprehensive time series analysis, DSGE models, and advanced forecasting. Includes unit root tests, ARIMA/seasonal models, VAR/cointegration, impulse responses, nonlinear models, forecast evaluation, and mixed-frequency regression.
Panel Data & Causal Inference51Statios supports canonical panel data estimators (FE, RE, Hausman test, 2SLS-IV), advanced dynamic panel methods (Arellano-Bond, System GMM), and modern causal inference designs (DiD, RDD, synthetic control, staggered DiD, IPW, mediation analysis).
Multivariate, Factor & Latent-Variable Analysis35Covers exploratory and confirmatory dimensionality reduction, clustering, reliability, latent structure discovery, and item response modeling — from classical PCA/factor analysis through modern IRT, SEM, and categorical methods.
Survival, Bayesian, Meta-Analysis & Resampling68Statistical methods for survival analysis, Bayesian inference, meta-analytic pooling, and resampling-based inference. Includes Kaplan-Meier curves, Cox proportional-hazards regression, parametric and competing-risks models, Bayesian linear/logistic/hierarchical regression with MCMC diagnostics, fixed- and random-effects meta-analysis, and bootstrap/jackknife/permutation methods.
Machine Learning & Choice Models31Statios provides machine learning and choice modeling tools: supervised learning with explainability (Random Forest, Gradient Boosting, SVM, Decision Trees, neural networks); model evaluation (cross-validation, ROC/AUC, confusion matrix, feature importance, SHAP); and conjoint analysis (traditional, choice-based, hierarchical Bayes) for preference modeling.
Complex Survey & Missing Data18Statios provides complex survey design analysis and missing data handling. Covers design-based inference (Taylor linearization, replicate weights, Rao-Scott corrections) and advanced imputation (MICE, EM, PMM) with proper uncertainty propagation.
Data Management, Transforms & Workflow68Data management, transforms and workflow procedures in Statios: dataset loading and multi-dataset workflows, direct cell editing with undo/redo, spreadsheet-style formulas for derived variables, variable metadata configuration (types, labels, measurement levels), data transformations (compute, recode, categorize, standardize, filter, lag/lead, dummy), data manipulation (merge, reshape, aggregate, collapse, append, sort, subset, deduplicate, rank, transpose, split/concat columns, pivot), data-to-table formatting, and session/macro recording for reproducible workflows.
AI, Automation, Graphics & Reporting19AI-powered statistical guidance, automated analysis pipelines, publication-quality graphics and report generation, plus a visual rule-based workflow recommender (Guided Path) and a node-based pipeline editor (RLab).

Available, with known limitations

These work today, with a caveat worth knowing.

CapabilityAreaLimitation
Regression Discontinuity (Sharp RDD)Regression & Generalized Linear ModelsBandwidth selection is ad-hoc (1.5*SD rule), not data-driven MSE-optimal.
Extended Regression Model — Probit (ERM-Probit)Regression & Generalized Linear ModelsIMR lambda SE not corrected for generated-regressor problem (Murphy-Topel correction not implemented).
Extended Regression Model — Ordered (ERM-Ordered)Regression & Generalized Linear ModelsLambda SE uncorrected; no p-values reported due to generated-regressor issue.
Treatment effect contrasts (ATE/ATT/ATC)Post-Estimation & DiagnosticsUses OLS-based potential outcomes, not designed for instrumental variables.
Diffusion-index direct forecast (user-supplied factors)Time Series & MacroeconometricsNOT a true Stock-Watson Dynamic Factor Model: there is no latent-factor extraction (no PCA / Kalman / state-space estimation of common factors inside the method) — the factors must be supplied as columns. Automatic latent-factor extraction is a separate upcoming feature.
Unrestricted MIDAS regressionTime Series & MacroeconometricsCurrently uses same dataset for both frequencies (pending multi-dataset resolution).
Almon polynomial MIDASTime Series & MacroeconometricsUses single dataset for both frequencies (pending multi-dataset support).
Beta polynomial MIDASTime Series & MacroeconometricsSingle-dataset limitation; multi-dataset pending (Phase 18).
Exponential Almon MIDASTime Series & MacroeconometricsSingle dataset; multi-dataset architecture pending.
MIDAS nowcastingTime Series & MacroeconometricsSingle-dataset implementation; true multi-frequency data awaits Phase 18.
MIDAS model comparisonTime Series & MacroeconometricsSingle-dataset; intended for separate high/low datasets pending multi-dataset manager.
Real Business Cycle modelTime Series & MacroeconometricsLog-linearized first-order approximation only; no occasionally binding constraints.
New Keynesian 3-equation modelTime Series & MacroeconometricsSimplified linearized form; no ELB, expectations-driven dynamics approximate.
DSGE calibrationTime Series & MacroeconometricsMoment-matching heuristic only; not Bayesian estimation or SMM.
Regression Discontinuity (Sharp RDD)Panel Data & Causal InferenceBandwidth selection is ad-hoc (not Imbens-Kalyanaraman or Calonico-Cattaneo-Titiunik data-driven optimal); local-linear only.
Propensity Score Matching (PSM)Panel Data & Causal InferenceUses naive paired SE; does not apply Abadie-Imbens (2006) variance correction for estimated scores and matching with replacement.
Hierarchical Agglomerative ClusteringMultivariate, Factor & Latent-Variable AnalysisO(n²) memory constraint; refuses n > 20,000 rows.
IRT Fit Indices (Approximate)Multivariate, Factor & Latent-Variable AnalysisApproximate method: fixed-1PL reference and diagonal weight matrix, not full Maydeu-Olivares M2.
Shared Frailty Cox ModelSurvival, Bayesian, Meta-Analysis & ResamplingGamma frailty only; variance estimated via martingale residuals.
Bayesian Hierarchical ModelSurvival, Bayesian, Meta-Analysis & ResamplingFallback uses OLS-shrinkage approximation without full MCMC; ICC computed from variance components.
Bayes FactorSurvival, Bayesian, Meta-Analysis & ResamplingBIC-based approximation; not exact marginal likelihood.
SHAP ValuesMachine Learning & Choice ModelsTreeExplainer unavailable in browser (WASM); fallback to KernelExplainer (sampled Shapley values, not exact TreeSHAP).
CHAIDMachine Learning & Choice ModelsStub implementation; returns placeholder.
QUESTMachine Learning & Choice ModelsStub implementation; returns placeholder.
C5.0Machine Learning & Choice ModelsStub implementation; returns placeholder.
Ensemble Trees ComparisonMachine Learning & Choice ModelsStub implementation; returns placeholder.
Tree ExportMachine Learning & Choice ModelsStub implementation; returns placeholder.
Compare TreesMachine Learning & Choice ModelsStub implementation; returns placeholder.
Hierarchical Bayes CBCMachine Learning & Choice ModelsEmpirical-Bayes / EM-style scheme, NOT Gibbs/Metropolis; samples are posterior modes, not draws; heterogeneity SD understates posterior uncertainty.
Session UndoData Management, Transforms & WorkflowSession-level undo not fully wired; edit.undo handles cell-level reversals.
Session RedoData Management, Transforms & WorkflowSession-level redo not fully wired; edit.redo handles cell-level reversals.
Large-dataset/columnar mode: bounded analysis suitedataWASM in-browser path capped at 1.5M rows or 900MB. CSV files up to 50M rows route to columnar store (Parquet/NPY), but only subset of analyses support column-on-demand (streaming). Binary formats (.xlsx/.sav/.dta) cannot be streamed.
WASM/Pyodide compatibility: numba and SHAP slower without JITmlWeb build stubs numba (Pyodide lacks LLVM JIT) and SHAP's _kernel_lib. Code runs interpreted (correct numerically, ~2–5× slower than compiled). KernelExplainer works; TreeExplainer unavailable in WASM.
Charts: 4 types shipped, 31 roadmap items incompletegraphsImplemented: histogram, scatter, boxplot, bar. Missing: violin, density, Q–Q, interaction, forest, ROC, funnel, survival curves, residual diagnostic panels, and 21 others from Phase 15 roadmap.
Extended regression diagnostics: some estimators in pre-release rigor improvementsregressionArellano-Bond, GMM, DSGE solvers present but some advanced econometric estimators have rigor enhancements on unmerged branches. Shipped versions work; see code docstrings for limitations (e.g., 'simplified solver', 'bounded lag range').

In active development

ItemAreaNotes
Statios web pivot: Python engine to Pyodide/WASMinfrastructureW1–W4 phases (June 2026): migrate sidecar to in-browser Pyodide via WorkerTransport. Desktop Tauri shell frozen. PWA deployment, OPFS file handling, web deploy pipeline. Completed handler count remains 445; no new features, transport refactor only.
Charts Expansion: 31 missing chart typesgraphsViolin, density (KDE), ridgeline, Q–Q, P–P, interaction plot, mosaic, ACF/PACF, IRF with CI, residual diagnostics (4-panel), forest/funnel, survival, ROC, Bland-Altman, coefficient, marginal effects, control chart, caterpillar. Phase 15 completion.
AI Orchestrator & Multi-Model RoutingaiPost-W4: intelligent provider selection, prompt caching, fallback logic, cost tracking. Statios relay authentication for managed web tier.
Quality backlog: UI refinements, performance, test coveragepolishPost-W4: Monaco editor for scripts, keyboard shortcuts, undo/redo on data transforms, canvas DataGrid for >10k rows, performance audits on large-dataset analyses.

On the roadmap

ItemAreaNotes
Advanced causal inference: bounds & sensitivity analysescausalLee bounds, Rosenbaum sensitivity, E-values. Handlers present (causal.lee_bounds, etc.) but backlog for UI polish and help content.
Exotic time-series models: TAR/STAR, NARDL, nowcastingtimeseriesHandlers exist (ts.tar, ts.nardl, ts.nowcasting) but limited test coverage and optional dependencies (ruptures for Bai–Perron). Deferred to post-quality phase.
Multivariate: MANOVA, discriminant, canonical correlation analysisHandlers registered (test.manova, analysis.discriminant, analysis.canonical_correlation) but frontend UI tabs not yet exposed in AnalysisModal. Feature-complete backend; UI routing needed.
Bayesian posterior predictive checks & Bayes factorsbayesHandlers: bayes.ppc, bayes.credible_interval, bayes.prior_posterior, bayes.bayes_factor. Deps (PyMC, ArviZ) optional. Lower priority; posterior work focuses on MCMC diagnostics.
Optimal scaling & ERM: categorical regression (CATREG), optimal transformsanalysisHandlers: analysis.catreg, analysis.catpca, analysis.overals, analysis.catcorr. Phase 23 backends exist but semi-automated UI guidance and diagnostic plots deferred.

Appendix — full procedure list

Descriptive Statistics & Hypothesis Tests 37

Descriptive statistics and statistical hypothesis tests covering parametric (t-tests, ANOVA), non-parametric (Mann-Whitney, Kruskal-Wallis, Friedman), correlation, and categorical association tests.

ProcedureWhat it doesStatus
Descriptive statisticsMean, median, variance, skewness, kurtosis, quantiles for numeric variables.Full
Frequency distributionFrequency tables with counts, percentages, cumulative percentages for categorical/numeric variables.Full
One-sample t-testTest if sample mean differs from hypothesized value; reports t-statistic, 95% CI, Cohen's d.Full
Independent samples t-testCompare means of two independent groups; includes Levene's test, optional Welch correction.Full
Paired t-testCompare means of two related/paired samples; reports difference mean, SE, 95% CI, effect size.Full
One-way ANOVACompare means across 3+ groups; includes Levene's test, Tukey HSD post-hoc, eta/omega-squared.Full
Welch's ANOVAHeteroscedastic one-way ANOVA without equal-variance assumption; Games-Howell post-hoc.Full
Two-way ANOVATest main effects and interaction of two factors; includes partial eta-squared effect sizes.Full
Repeated measures ANOVAWithin-subjects ANOVA for 2+ repeated measures; Mauchly sphericity, GG/HF corrections.Full
Yuen's trimmed t-testRobust two-group t-test using 20% trimmed means; insensitive to heavy tails and outliers.Full
Mixed ANOVASplit-plot design: between-subject factor and within-subject repeated measures.Full
Dunnett's testPost-hoc test comparing each treatment group to a control group.Full
Cochran-Armitage testTrend test for proportions across ordered groups.Full
Mann-Whitney U testNon-parametric two-group test; exact/asymptotic p-values, rank-biserial effect size.Full
Wilcoxon signed-rank testNon-parametric paired test; exact/asymptotic p-values, effect size r.Full
Kruskal-Wallis H testNon-parametric 3+ group test; Dunn post-hoc with Bonferroni correction.Full
Friedman testNon-parametric repeated measures test; Kendall's W effect size.Full
Chi-square test of independenceAssociation between two categorical variables; includes Cramér's V, Yates correction disclosure.Full
Chi-square goodness-of-fitTest if observed distribution matches expected frequencies.Full
Fisher's exact testExact test for 2x2 contingency table; alternative: one/two-sided.Full
Pearson correlation matrixPairwise Pearson/Spearman/Kendall correlations with p-values; listwise NA handling.Full
Correlation (pairwise)Single correlation with confidence interval (Fisher z-transform for Pearson).Full
Partial correlationCorrelation between two variables controlling for others; corrected df.Full
Effect size calculatorCohen's d/Hedges' g, eta/omega-squared, Cramér's V, Fisher's f-squared.Full
Power analysisSample size / power calculation for t-test, ANOVA, correlation, chi-square, regression.Full
Sign testNon-parametric paired/one-sample test; binomial test on signs of differences.Full
Runs testWald-Wolfowitz test for randomness; dichotomize by median/mean or threshold.Full
McNemar testChange detection in paired binary outcomes; 2x2 table symmetry.Full
Cochran Q testNon-parametric repeated measures for binary outcomes (3+ measurements).Full
Jonckheere-Terpstra testNon-parametric alternative to ANOVA for ordered groups (monotonic trend).Full
Kendall's WCoefficient of concordance for 2+ judges/raters across multiple items.Full
Kendall's tauRank correlation between two variables (tau-b for ties).Full
Mood's median testNon-parametric k-group test based on median; robust to outliers.Full
Permutation testPermutation-resampling test (10000 permutations); mean difference or custom statistic.Full
Kernel density estimateNon-parametric density estimation; Scott/Silverman bandwidth, Gaussian kernel.Full
MANOVAMultivariate analysis of variance; Wilks' lambda, Pillai, Hotelling-Lawley, Roy's root.Full
Distribution fitting + GOFFit Exponential-Family and related distributions (gamma, exponential, lognormal, Weibull, normal) by MLE, rank by AIC, and test fit with Kolmogorov-Smirnov. Matches R MASS::fitdistr + ks.test.Full

Regression & Generalized Linear Models 142

Statios's model namespace covers 50+ regression and generalized linear estimation procedures, from OLS to latent-variable extensions. Methods range from standard parametric models (linear, logistic, count) to specialized estimators for endogenous/selected samples, censoring, and complex errors. All predictor-taking handlers auto-expand categorical predictors R-style (labeled indicator columns, named reference level, factors field) — no manual dummy coding required.

ProcedureWhat it doesStatus
Linear Regression (OLS)Standard ordinary least squares with diagnostics (R², F-test, VIF, Durbin-Watson, residual normality). Categorical predictors are auto-expanded R-style (labeled indicators, named reference level, factors field) — no manual dummy coding required.Full
Linear Regression — Robust SEs (HC0–HC3)OLS with heteroskedasticity-consistent standard errors; selectable HC0 (White), HC1 (Stata default), HC2, or HC3 (MacKinnon & White 1985).Full
Linear Regression — Newey-West HAC SEsOLS with Newey-West heteroskedasticity- and autocorrelation-consistent (HAC) standard errors (Bartlett kernel, set hac_lags); standard for time-series regressions.Full
Errors-in-Variables RegressionMethod-of-moments correction for classical measurement error (attenuation bias) using per-regressor reliability ratios; reports naive OLS vs corrected coefficients.Full
SIMEX (measurement-error correction)Corrects attenuation bias from a regressor measured with known additive error variance by simulation-extrapolation: adds increasing extra noise, refits OLS at each lambda (averaging over B Monte-Carlo replicates), and extrapolates the coefficient quadratically back to lambda=-1 to recover the error-free estimate.Full
Frisch-Waugh-Lovell Partial SlopePartials the controls out of both Y and the focal regressor and regresses the residuals through the origin; numerically equals the focal coefficient in the full OLS, with the partial correlation.Full
Omitted-Variable Bias DecompositionDecomposes omitted-variable bias for each included regressor into short-model vs long-model coefficients and the exact gamma*delta contribution of each omitted variable.Full
Log-Y Retransformation (Duan Smearing)Fits a log-scale OLS and reports Duan's smearing factor S=mean(exp(residuals)) plus naive vs bias-corrected (smeared) retransformed predictions on the original scale.Full
Potential Outcomes (ATE/ATT/ATU)Reports the average treatment effect (ATE), the effect on the treated (ATT) and on the untreated (ATU) in Rubin's counterfactual notation by regression adjustment under unconfoundedness: an OLS fit of the outcome on treatment, covariates and their interactions predicts each unit's pair (Y(1), Y(0)), with delta-method or bootstrap standard errors.Full
Selection Model (MNAR missingness)Heckman (1979) two-step selection model for non-ignorable (MNAR) missingness: a probit models whether the outcome is observed, the outcome equation is corrected for selection on unobservables via the inverse Mills ratio, and rho (with its rho=0 ignorability test) quantifies the selection correlation, reported alongside naive complete-case OLS for comparison.Full
Pattern-Mixture Model (MNAR)Pattern-mixture model (Little 1993) for non-ignorable (MNAR) missing outcomes: partitions the sample by the outcome's missingness, fits the completers' regression, imputes the missing outcomes under a sensitivity restriction whose delta shifts the missing-group mean (delta=0 = MAR), and reports the pattern-averaged mean, fraction missing, and a sensitivity curve.Full
Continuous-Treatment Dose-Response (GPS)Estimates the average dose-response function of a continuous treatment with the Hirano-Imbens generalized propensity score: models the treatment given covariates as Normal, regresses the outcome on a flexible function of the treatment and the GPS, and averages to trace E[Y(t)] over a grid of doses, returning the dose-response curve, the GPS-adjusted average marginal effect, and the naive unadjusted slope for comparison.Full
Abadie Kappa-Weighting (compliers)Uses Abadie's (2003) kappa weights kappa = 1 - D(1-Z)/(1-p(X)) - (1-D)Z/p(X) to estimate the complier share, the complier mean outcome, the mean covariates among compliers versus the full sample, and a kappa-weighted LATE in a binary-instrument / binary-treatment IV setting.Full
Marginal Treatment Effect (MTE)Estimates the Heckman-Vytlacil marginal treatment effect via local IV / generalized Roy: fits the propensity score and a flexible control function in p, differentiates to trace the MTE over the unobserved-resistance dimension, integrates it with the appropriate weights to recover ATE/ATT/ATUT/LATE, and tests for essential heterogeneity.Full
Synthetic DiDSynthetic Difference-in-Differences (Arkhangelsky et al. 2021) combines synthetic-control unit weights with difference-in-differences time weights to estimate the ATT on a balanced treated/control panel, reporting the SDID effect, a placebo/jackknife standard error, the sparse non-negative unit weights and the pre-treatment fit.Full
Changes-in-Changes (Athey-Imbens)Nonlinear generalization of difference-in-differences (Athey & Imbens 2006) that maps the treated-pre outcome distribution through the control group's pre-to-post distributional change to identify the entire counterfactual treated distribution, reporting the ATT plus quantile treatment effects (0.25/0.5/0.75) that reveal heterogeneous effects ordinary DiD would miss.Full
Doubly-Robust DiD (Sant'Anna-Zhao)Estimates the ATT in a 2x2 difference-in-differences design by combining an outcome-regression DiD with an inverse-propensity-weighted DiD into the doubly-robust estimator of Sant'Anna & Zhao (2020), which stays consistent if either the propensity-score model or the untreated-outcome model is correct, reports influence-function standard errors with the IPW and OR components and the simple DiD for contrast, and reduces exactly to the canonical mean 2x2 DiD when no covariates are supplied.Full
Honest DiD (parallel-trends sensitivity)Rambachan & Roth (2023) sensitivity analysis for difference-in-differences: estimates the event-study coefficients, then bounds the post-treatment effect under a relative-magnitudes restriction (Mbar) on how far post-treatment trend violations can exceed observed pre-trends, reporting robust confidence sets across Mbar and the breakdown Mbar at which significance is lost.Full
Sun-Abraham Event StudyInteraction-weighted (IW) event-study estimator that saturates the model with cohort-by-relative-period interactions and aggregates them by cohort shares, recovering the true dynamic treatment path (and clean pre-trends) under staggered adoption with heterogeneous effects where conventional dynamic TWFE is contaminated.Full
Callaway-Sant'Anna Staggered DiDCallaway & Sant'Anna (2021) staggered difference-in-differences estimating group-time average treatment effects ATT(g,t) via clean 2x2 DiD against not-yet-treated/never-treated comparison units, aggregated to an overall ATT, an event-study (dynamic) profile, and per-cohort effects, with a naive two-way FE benchmark to expose heterogeneous-timing bias.Full
Causal Forest (heterogeneous effects)Honest generalized random forest (Wager-Athey) that estimates the conditional average treatment effect tau(x) via split-sample honest trees with R-learner local centering, returning the ATE with a forest-based SE, the distribution of individual CATEs, and the leading covariate driving effect heterogeneity.Full
Double / Debiased MLEstimates a debiased treatment effect in the partially-linear model Y = theta*D + g(X) + e by cross-fitting gradient-boosted nuisance regressions for E[Y|X] and E[D|X] and solving the orthogonal (Neyman) moment on the residuals, reporting theta with robust SE/CI alongside the biased naive single-regression estimate for contrast.Full
McCrary Density Test (RD manipulation)McCrary (2008) density-discontinuity test that fits separate triangular-kernel local-linear density estimates on each side of an RD cutoff and tests whether the log-density of the running variable jumps at the threshold, detecting sorting/manipulation that would invalidate the RD design.Full
Propensity-Score SubclassificationEstimates the ATE and ATT of a binary treatment by fitting a propensity-score logit, splitting units into quantile strata of the score, and combining the within-stratum treated-minus-control mean differences weighted by stratum size (ATE) or treated share (ATT), with per-stratum effects and a stratified standard error.Full
Mahalanobis Matching (ATT)Nearest-neighbour matching estimator of the Average Treatment effect on the Treated that matches each treated unit to its closest control(s) in Mahalanobis covariate distance (pooled-control inverse covariance), imputes the counterfactual from the matched controls, and reports the ATT with an approximate Abadie-Imbens standard error and covariate balance.Full
Box-Cox Transformed RegressionEstimates the Box-Cox power-transform lambda maximizing the concentrated profile log-likelihood, then fits OLS on the transformed (strictly-positive) response.Full
Stochastic Frontier AnalysisMaximum-likelihood Aigner-Lovell-Schmidt stochastic frontier (half-normal inefficiency) for production or cost functions, returning frontier coefficients, sigma_v/sigma_u, lambda, gamma, mean technical efficiency E[exp(-u)] and per-observation efficiency scores.Full
Finite-Mixture Regression (latent class)Fits a finite mixture of K linear regression regimes by EM, recovering each latent class's own coefficient vector, error variance, and mixing proportion together with the log-likelihood, BIC, and posterior class-assignment summary.Full
Dirichlet-Process Mixture DensityBayesian nonparametric density estimation that fits an infinite (Dirichlet-process) Gaussian mixture via truncated stick-breaking, automatically inferring the effective number of mixture components and returning their means, variances, weights, and the estimated density curve.Full
Finite-Mixture Count (latent class)Fits a finite mixture of K latent-class Poisson regressions by EM, recovering each regime's own log-link coefficient vector and mixing proportion together with the log-likelihood, BIC/AIC, and posterior class-assignment summary to capture unobserved heterogeneity and overdispersion in count data.Full
Simultaneous Quantile RegressionFits linear quantile regressions at several quantiles at once and uses a joint paired (xy) bootstrap covariance to report per-quantile coefficient tables and to test whether slopes change across the conditional distribution (e.g. tau=0.25 vs 0.75).Full
Quantile Regression ForestEstimates full conditional quantiles (Meinshausen 2006) with a random forest by reading the weighted empirical response distribution in each leaf, revealing heteroskedastic conditional intervals that widen with predictors plus an out-of-bag interval-coverage check and variable importance.Full
Additive Quantile Regression (quantile GAM)Fits a tau-quantile regression in which each predictor enters through a smooth cubic B-spline term, minimizing the asymmetric check-loss via linear programming to recover nonlinear conditional quantile curves.Full
Composite Quantile RegressionEstimates a single shared slope vector by pooling the check-loss across an equally-spaced grid of quantiles (with quantile-specific intercepts), giving slope estimates that match OLS under Gaussian errors and beat it in efficiency under heavy-tailed errors, with bootstrap standard errors.Full
Expectile RegressionNewey-Powell asymmetric-least-squares expectile regression, fit by iteratively reweighted least squares to estimate any conditional tau-expectile (tau=0.5 reproduces OLS exactly), with heteroskedasticity-consistent sandwich standard errors and an asymmetry interpretation.Full
Quantile Count Regression (jittering)Estimates conditional quantiles of an overdispersed count response by the Machado-Santos Silva (2005) jittering method -- uniform noise is added to the integer counts, quantile regression is run on the log-transformed scale, and coefficients are averaged over many jitter draws to give stable, monotone-in-tau multiplicative quantile effects.Full
Generated-Regressor Inference (Murphy-Topel)Two-step estimation that regresses a stage-2 outcome on the fitted values from a first-stage OLS, reporting both the naive stage-2 standard errors and the Murphy-Topel/Pagan-corrected standard errors that propagate the first-stage estimation uncertainty into the generated regressor.Full
Heteroskedastic ProbitHarvey (1976) binary probit in which the latent-error standard deviation is modelled as sigma_i = exp(z_i'gamma), estimated by BFGS maximum likelihood, returning separate mean (beta) and variance (gamma) coefficient tables plus a likelihood-ratio test of homoskedasticity (H0: gamma=0).Full
Kernel Regression (Nadaraya-Watson)Nonparametric Nadaraya-Watson local-constant kernel smoother estimating the conditional mean of one outcome on one predictor (Gaussian or Epanechnikov kernel, Silverman's rule-of-thumb or user bandwidth), returning fitted values, a 100-point smooth curve, an R-squared-like fit measure, and the bandwidth used.Full
Kernel Conditional DensityEstimates the full conditional density f(y|x) with a double Gaussian kernel (Nadaraya-Watson weighting in x combined with a kernel in y), returning conditional 0.1/0.5/0.9 quantile curves and the conditional mean as functions of x so heteroskedastic spread, skew and multimodality are visible where a mean regression cannot show them.Full
Single-Index Model (Ichimura)Semiparametric single-index regression y = g(x'beta) + e estimated by Ichimura's semiparametric least squares, recovering the index direction beta (identified up to scale, unit-norm normalized) by minimizing the leave-one-out kernel-regression residual sum of squares while leaving the link g(.) fully nonparametric, and reporting standardized and raw-scale index coefficients, the data-driven bandwidth, a leave-one-out pseudo-R-squared, and the estimated link curve.Full
Maximum Score (Manski)Manski's semiparametric maximum-score estimator for binary choice, recovering the scale-normalized coefficient direction by maximizing correctly-predicted signs of x'beta under only a conditional-median error restriction, staying consistent where logit/probit fail under heteroskedasticity or non-normal errors.Full
Translog Production/Cost FunctionEstimates a second-order flexible (transcendental logarithmic) production or cost function by OLS on log-inputs, their squares and pairwise cross-products, reporting output elasticities at the sample mean, returns-to-scale, and an F-test of the Cobb-Douglas (no-second-order-terms) restriction.Full
Empirical Likelihood (mean)Owen's nonparametric empirical-likelihood inference for a population mean or mean vector, returning the -2 log EL-ratio statistic with its chi-square p-value plus a distribution-free EL confidence interval that requires no normality assumption.Full
Copula Fitting (bivariate dependence)Fits Gaussian, Clayton, Gumbel and Frank copulas to two variables' rank pseudo-observations by maximum likelihood, reporting each family's parameter, log-likelihood and AIC, the AIC-selected copula, and Kendall's tau and Spearman's rho.Full
Gaussian-Process RegressionNonparametric Bayesian regression with an RBF + white-noise kernel that learns its length-scale and noise level by maximizing the log-marginal-likelihood, returning a smooth posterior mean plus per-point predictive uncertainty.Full
BART (Bayesian Additive Regression Trees)Flexible nonparametric regression that models the conditional mean as a regularized sum of trees fit by Bayesian backfitting MCMC, returning posterior-mean fitted values, predictive uncertainty, variable-inclusion importance, and in-sample R-squared.Full
Gaussian-Process ClassificationFits a Gaussian-process classifier (RBF kernel, Laplace approximation) to a binary 0/1 outcome, reporting in-sample accuracy/log-loss/AUC, learned kernel hyperparameters, and a predicted-probability summary to capture smooth nonlinear class boundaries a linear logit would miss.Full
k-Nearest-Neighbours RegressionNonparametric regression that predicts each outcome as the average of its k nearest neighbours in standardized (Euclidean) predictor space, reporting an in-sample fit, a leave-one-out cross-validated RMSE and CV R-squared, and the chosen k.Full
Spline Regression (B-splines)Fits a flexible nonlinear curve by building a B-spline basis of one predictor (patsy bs() with given degrees of freedom and polynomial degree) and running OLS of the response on that basis, returning the fitted curve on a 100-point grid, the data-quantile knot locations, the basis dimension, R-squared and residual MSE.Full
Generalized Additive Model (splines)Fits y = intercept + sum_j f_j(x_j) + e with each term a penalized cubic B-spline (statsmodels GLMGam, Gaussian family), selecting the smoothing penalty by GCV and reporting per-term effective degrees of freedom, deviance, pseudo-R^2, and a fitted-value summary.Full
GAM (Poisson / Logit)Generalized additive model for count (Poisson/log) or binary (Binomial/logit) outcomes that fits each predictor as a penalized cubic B-spline smooth via penalized IRLS with GCV/UBRE smoothing selection, reporting per-smooth effective degrees of freedom, deviance explained, fitted curves with confidence bands, and approximate significance.Full
Local-Linear RegressionNonparametric degree-1 local-polynomial kernel smoother that fits a kernel-weighted least-squares line at each point and takes its intercept as the conditional-mean estimate (Gaussian or Epanechnikov kernel, Silverman rule-of-thumb or user bandwidth), correcting the boundary bias of Nadaraya-Watson and returning fitted values, a 100-point smooth curve, effective degrees of freedom, and an R-squared-like fit measure.Full
Local Polynomial RegressionNonparametric kernel-weighted local polynomial smoother of arbitrary degree that generalizes Nadaraya-Watson (degree 0) and local-linear (degree 1) fits, returning the fitted curve, its estimated first derivative, and a smoother effective-df fit measure, with higher degrees reducing boundary bias.Full
Isotonic Regression (PAVA)Nonparametric monotone least-squares fit of a response on a single predictor via the pool-adjacent-violators algorithm, returning the fitted step function (level sets), R^2, and the imposed direction.Full
Wavelet Regression (denoising)Nonparametric denoising of a response treated as a signal ordered by a predictor via discrete wavelet transform with soft-thresholding (VisuShrink universal or hybrid SureShrink), returning the denoised fitted curve, robust MAD noise estimate, retained-coefficient sparsity, and reconstruction diagnostics.Full
Wavelet CoherenceComputes squared wavelet coherence R^2 in [0,1] between two time series across time and scale via a hand-rolled FFT-based Morlet continuous wavelet transform, smoothed cross-wavelet and power spectra (Torrence & Compo 1998; Grinsted et al. 2004), summarizing the dominant common period, coherence by short/medium/long scale bands, and the cone-of-influence-masked global coherence.Full
Partially Linear (Robinson)Robinson (1988) sqrt(n)-consistent semiparametric partially-linear regression y = x'beta + g(z) + e, estimating the linear slopes beta by the double-residual method (Nadaraya-Watson partialling-out of y and each x on z, then residual-on-residual OLS) while leaving g(z) fully nonparametric.Full
Overdispersion Test (Cameron-Trivedi)Cameron-Trivedi (1990) regression-based test of the Poisson equidispersion assumption: fits a Poisson GLM, then runs auxiliary through-origin OLS of the studentised squared-residual statistic on g(mu) for g(mu)=mu (NB1) and g(mu)=mu^2 (NB2), reporting per-variant alpha-hat, a one-sided t-test of H0 alpha=0, and an overdispersion verdict.Full
Conway-Maxwell-Poisson RegressionCOM-Poisson GLM (log link on lambda) with a jointly-estimated dispersion parameter nu that flexibly handles under- (nu>1) and over-dispersed (nu<1) counts, fit by maximum likelihood with a stable truncated normalizing constant and a Wald test of nu=1 against the Poisson.Full
Endogenous Count (control function)Poisson count regression with an endogenous regressor corrected by two-stage residual inclusion (control function): a first-stage OLS residual on instruments is added to the count model, yielding the endogeneity-corrected effect, a control-function residual coefficient that doubles as an endogeneity test, and bootstrap standard errors.Full
System OLS (multi-equation)Estimates a system of linear equations equation-by-equation by OLS (intercept added), reporting per-equation coefficient tables (coef/se/t/p/CI), R2 and n, plus the cross-equation residual covariance Sigma (symmetric, PSD) and residual correlation on the common listwise-complete sample - the SUR/FGLS baseline.Full
Joint Confidence RegionFits OLS and reports the joint F test that a chosen set of coefficients is simultaneously zero, the exact confidence-ellipse geometry (centre, covariance submatrix, eigen-axes and half-axis lengths) for exactly two coefficients, and Bonferroni simultaneous intervals, contrasting the tilted exact region with the conservative axis-aligned box.Full
General GLS (specified Omega)Feasible generalized least squares (Aitken) under a user-specified error covariance: AR(1) Toeplitz errors, groupwise heteroskedasticity (one variance per group level), or power heteroskedasticity (var proportional to |fitted|^delta), reporting the estimated structure parameter and an efficient coefficient table.Full
Complementary Log-Log RegressionBinary regression with the asymmetric complementary log-log link P(y=1)=1-exp(-exp(x'beta)) for rare-event data and grouped/discrete-time proportional-hazards (survival) models, returning a coef/se/z/p/CI table with log-likelihood and AIC/BIC.Full
GEE (Population-Averaged)Generalized estimating equations fit a population-averaged (marginal) regression for clustered or longitudinal data, modelling within-cluster dependence through a working correlation (independence / exchangeable / AR-1) while reporting robust sandwich standard errors valid even under correlation misspecification.Full
LIML (Limited-Information ML)Limited-information maximum likelihood IV estimator that is less biased than 2SLS under weak or many instruments, reporting the k-class kappa eigenvalue, robust coefficient table with confidence intervals, first-stage instrument-strength diagnostics, and an overidentification test.Full
Two-Sample IV (TS2SLS)Two-sample two-stage least squares: estimates a single endogenous regressor's structural coefficient by fitting the first stage in one half of the data and carrying the fitted endogenous value into the reduced-form/structural equation in the other half (Inoue-Solon TS2SLS; Angrist-Krueger TSIV design), reporting the coefficient with a robust SE, first-stage strength, and sample sizes.Full
Semiparametric IV (sieve)Newey-Powell / Ai-Chen series (sieve) two-stage instrumental-variables estimation of an unknown dose-response g(endog) approximated by a polynomial sieve and instrumented by a sieve basis of the instruments, returning the estimated g(.) curve on a grid, the average derivative, HC1-robust standard errors, first-stage relevance, and a naive sieve-OLS comparison to expose endogeneity bias.Full
Lewbel Heteroskedasticity IVEstimates a regression with an endogenous regressor when no external instruments exist, by constructing Lewbel (2012) generated instruments from heteroskedasticity in the first-stage residual and running heteroskedasticity-robust 2SLS, reporting the first-stage F and an OLS-vs-Lewbel comparison.Full
Generalized Empirical Likelihood (GEL)Information-theoretic estimation of a linear IV / moment-condition model via the GEL saddle-point (empirical likelihood, exponential tilting, or continuous-updating), returning coefficients with sandwich SEs, an LR-type overidentification test, and nonnegative implied probabilities; reproduces 2SLS exactly when just-identified.Full
Empirical-Likelihood RegressionEstimates linear-regression coefficients on the OLS estimating-equation moments E[x(y - x'beta)]=0 and tests/intervals them via Owen's empirical likelihood, giving distribution-free, generally-asymmetric EL confidence intervals plus chi-square-calibrated EL-ratio p-values whose point estimates coincide with OLS.Full
Continuously-Updated GMM (CUE)Continuously-updated GMM for a linear IV model that re-estimates the optimal weight matrix as a function of the parameters inside the objective and minimises jointly, returning CUE coefficients with robust standard errors, z/p-values, the Hansen J overidentification test, and (when just-identified) an exact 2SLS match.Full
Optimal-Instrument GMMEfficient GMM for a linear IV model that builds Chamberlain (1987) optimal instruments A(z)=E[x|z]/Var(e|z) from a flexible projection of the regressors on the instruments plus a fitted conditional-variance model, reporting beta with efficient-GMM standard errors, first-stage fit, Hansen's J overidentification test, and an efficiency comparison versus plain 2SLS.Full
Nonlinear GMM (exponential moments)Two-step efficient GMM for the exponential / multiplicative-error model y = exp(x'beta)*eta with instruments, using Mullahy's multiplicative moment E[z*(y*exp(-x'beta)-1)]=0 to identify possibly-endogenous regressors (a nonlinear analogue of 2SLS) and reporting the coefficient table plus Hansen's J overidentification test.Full
Method of Simulated MomentsMethod of Simulated Moments / indirect inference for a univariate AR(1)-plus-measurement-noise model whose likelihood is intractable: parameters (phi, sigma_eta, sigma_e) are chosen so simulated auxiliary statistics (variance and first two autocovariances) match the data's, minimizing a scale-free weighted distance with common random numbers and reporting estimates, sandwich SEs with the (1+1/S) simulation-noise inflation, the objective value, and the moment-match table.Full
Minimum Distance (CMD)Classical minimum-distance (minimum-chi-square) estimation that recovers a structural IV slope from the OLS reduced-form coefficients by minimizing (pi_hat - h(theta))' W (pi_hat - h(theta)) with the efficient weight W = inv(Var(pi_hat)), reporting theta_hat, CMD standard errors, the overidentification chi-square, and (in the just-identified case) exact agreement with 2SLS.Full
IV Quantile Regression (Chernozhukov-Hansen)Estimates the structural effect of an endogenous regressor on a chosen conditional quantile of the outcome using one or more instruments, via the Chernozhukov-Hansen inverse-quantile-regression grid search that drives the instrument coefficient to zero, correcting the endogeneity bias that contaminates ordinary quantile regression.Full
Oaxaca-Blinder DecompositionDecomposes the mean outcome gap between two groups from separate within-group OLS fits into the threefold endowments/coefficients/interaction components and the twofold pooled-reference (Neumark) explained vs unexplained split.Full
Prais-Winsten AR(1) RegressionIterative AR(1) FGLS that quasi-differences the data and retains a scaled first observation, correcting OLS for first-order serial correlation; reports rho and before/after Durbin-Watson.Full
Linear Probability ModelOLS of a binary 0/1 outcome with HC1 robust SEs; coefficients are marginal effects on P(y=1), and the share of out-of-[0,1] fitted values is flagged.Full
Generalized Linear Model (family x link)Unified GLM: choose the exponential family (gaussian, poisson, binomial, gamma, inverse-Gaussian) and link (identity, log, logit, probit, inverse, sqrt). Reports coefficients, deviance, AIC/BIC.Full
Linear Combination of Coefficients (lincom)Tests an arbitrary linear combination g'β of OLS coefficients with its standard error, t, p-value, and 95% CI (Stata-style lincom).Full
Delta-Method (nonlinear combination)Delta-method standard error for a sum / difference / ratio / product of two OLS coefficients, with t, p-value, and 95% CI.Full
Restriction Test (F / Wald / LR)General joint-exclusion test: F, Wald, and likelihood-ratio statistics for dropping a subset of regressors (unrestricted vs restricted OLS).Full
Binary Logistic RegressionBinary outcome (0/1) modeled as probability; reports odds ratios, McFadden R², classification accuracy.Full
Probit RegressionBinary choice via normal CDF link; includes marginal effects and classification accuracy.Full
Multivariate ProbitJointly models 3+ correlated binary outcomes by maximum simulated likelihood with the GHK simulator, returning per-equation probit coefficients plus the full cross-equation error correlation matrix.Full
Bivariate ProbitJointly estimates two binary-outcome probit equations with correlated normal errors by maximizing the bivariate-normal log-likelihood, reporting both coefficient tables, the error correlation rho with a Wald rho=0 test that detects whether the outcomes can be modeled as two independent probits, the log-likelihood and n.Full
Ordered LogitOrdinal outcome; reports coefficients, threshold parameters, and McFadden pseudo-R².Full
Generalized Ordered LogitFits an ordinal-outcome model that relaxes the proportional-odds (parallel-lines) assumption by letting each cumulative split have its own ML-estimated coefficient vector, reporting per-threshold coefficient tables, a Brant-style Wald test (with an LR cross-check) of parallel lines, log-likelihood, AIC/BIC, and an interpretation.Full
Sequential / Continuation-Ratio LogitModels an ordinal outcome as a sequence of binary continuation-ratio logits -- one per ordered transition, fit on the at-risk subset {Y>=r} -- yielding stage-specific coefficients for outcomes whose categories are stages literally passed through in order.Full
Ordered ProbitOrdinal outcome via normal CDF; threshold parameters and pseudo-R² included.Full
Multinomial LogitMulti-category outcome; reports coefficients and relative-risk ratios per category relative to base.Full
Choice-Based Sampling (WESML)Corrects a conditional/multinomial logit for endogenous (choice-based) stratified sampling using the Manski-Lerman WESML estimator, reweighting each observation by population-share / sample-share of its chosen alternative to recover consistent population coefficients with robust sandwich standard errors, alongside a side-by-side comparison to the biased unweighted fit.Full
Multinomial Probit (GHK)Estimates a random-utility multinomial probit on long-format choice data by maximum simulated likelihood with the Geweke-Hajivassiliou-Keane (GHK) recursive simulator, recovering attribute coefficients and the identified correlated-error covariance while relaxing the IIA restriction of the conditional logit.Full
Nested LogitTwo-level nested logit for discrete choice estimated by FIML, relaxing IIA within nests via per-nest dissimilarity (lambda) parameters with a 0<lambda<=1 RUM-consistency check and a likelihood-ratio test against the conditional logit.Full
Mixed (Random-Parameters) LogitRandom-parameters (mixed) logit estimated by simulated maximum likelihood on long-format discrete-choice data, drawing Normal mixing coefficients via Halton draws to recover preference means and the std devs that quantify unobserved heterogeneity.Full
Maximum Simulated LikelihoodFits a random-coefficient (mixed) binary-choice probit by Maximum Simulated Likelihood, approximating the intractable integral over unobserved slope heterogeneity with shifted-Halton low-discrepancy draws and reporting the mean coefficients, the heterogeneity standard deviation, the simulated log-likelihood, the number of draws, and a likelihood-ratio comparison to the pooled (no-heterogeneity) probit.Full
Conditional Logit (McFadden)Estimates McFadden's choice-specific conditional logit on long-format discrete-choice data, recovering how alternative-varying attributes drive the probability that each alternative is chosen within a choice occasion, with a full coefficient table, log-likelihood, and McFadden pseudo-R2.Full
Poisson RegressionCount data (non-negative integer); reports incidence-rate ratios (IRRs) and dispersion diagnostics; optional exposure/offset term for rate (frequency) models.Full
Negative Binomial (NB1 / NB2)Overdispersed count data; MLE-estimated dispersion (alpha). Selectable NB2 (quadratic variance, default) or NB1 (linear variance). Includes likelihood-ratio test of Poisson vs NB.Full
Zero-Inflated Poisson (ZIP)Count data with excess zeros via mixing; separate count and zero-inflation equations with Vuong test.Full
Zero-Inflated Negative Binomial (ZINB)Overdispersed count data with excess zeros; count + inflation + alpha parameter estimated jointly.Full
Hurdle ModelTwo-part: logit for zero vs nonzero + truncated Poisson for positive counts.Full
Tobit Regression (Censored)Outcome censored at known bound (lower/upper); MLE estimation with sigma parameter.Full
Censored LAD (Powell CLAD)Powell's (1984) Censored Least Absolute Deviations estimator for left-censored (Tobit-type) outcomes, fit by an iterated median-regression linear-programming scheme with bootstrap standard errors, consistent under conditional-median restrictions and robust to heteroskedastic or heavy-tailed errors where Tobit is biased.Full
Truncated RegressionSample truncated (observations above threshold retained); MLE with sigma.Full
Heckman Selection Model (Two-Step)Endogenous sample selection via probit + IMR augmentation; reports rho correlation coefficient.Full
Endogenous Switching (Roy model)Estimates a two-regime Roy switching regression with regime-specific outcome equations and a probit selection equation, correcting each regime for selection on unobservables via its own inverse Mills ratio and reporting the rho_0/rho_1 error-correlation terms.Full
Beta RegressionOutcome in (0,1); precision parameter (phi) and mean-model coefficients estimated jointly.Full
Fractional-response quasi-MLE (Papke-Wooldridge)Outcome in [0,1]; Papke-Wooldridge (1996) Bernoulli quasi-MLE — GLM with binomial family + logit link; includes marginal effects.Full
Quantile RegressionConditional quantile (median or specified tau); pseudo-R² and quantile-specific inference.Full
Ridge Regression (L2)L2 regularized OLS on standardized predictors; alpha controls shrinkage intensity.Full
LASSO Regression (L1)L1 regularized OLS; automatic feature selection via coefficient sparsity. Standardized fit.Full
Elastic Net (L1+L2)Combined L1/L2 penalty on standardized predictors; l1_ratio balances penalties.Full
Elastic Net (cross-validated)Elastic-net penalized linear regression that standardizes predictors and uses K-fold cross-validation to pick the penalty strength lambda at minimum CV-MSE and at the 1-SE rule, reporting the CV-MSE curve, original-scale selected coefficients, the number selected, and R^2.Full
Robust Regression (Huber M-estimator)M-estimator resistant to outliers; asymptotic-normal Wald inference (z-stat); no t-distribution df.Full
Stepwise SelectionGreedy forward/backward/both; adds/removes by AIC or BIC until no improvement.Full
Weighted Least Squares (WLS)User-supplied or estimated weights (FGLS); corrects heteroscedasticity.Full
Feasible Generalized Least Squares (FGLS)Cochrane-Orcutt iterative GLSAR; corrects serial correlation in errors.Full
Nonlinear Least Squares (NLS)User-provided formula and starting values; scipy.optimize curve_fit minimizes residual sum of squares.Full
Seemingly Unrelated Regressions (SUR)Multiple equations fitted jointly accounting for cross-equation correlation; Breusch-Pagan test included.Full
Three-Stage Least Squares (3SLS)System of equations with endogenous right-hand-side variables; IV + SUR combined.Full
FIML (simultaneous system)Full-Information Maximum Likelihood joint estimation of a two-equation linear simultaneous-equations system, maximizing the system Gaussian log-likelihood concentrated over the residual covariance (with the ln|det(B)| Jacobian) to recover structural coefficients with FIML standard errors, the residual covariance/correlation, the system log-likelihood, and a side-by-side comparison against equation-by-equation 2SLS.Full
Generalized Method of Moments (GMM)IV-GMM with moment conditions; J-test of overidentifying restrictions.Full
Panel Fixed EffectsWithin-entity estimator; removes time-invariant heterogeneity. Clustered SEs on entity. Optional two-way (entity + time) effects.Full
Panel Random EffectsGLS random-effects estimator; assumes random effect uncorrelated with predictors.Full
Linear Mixed Model (Random Intercepts/Slopes)REML by default; fixed-effect Wald-z inference. AIC/BIC reported only under ML.Full
Intraclass Correlation (ICC)ICC(1) from one-way random-intercept model; variance-partition coefficient for clustering.Full
Difference-in-Differences (DiD)Treatment×Post interaction; cluster-robust or HC1 SEs depending on clustering.Full
Regression Discontinuity (Sharp RDD)Local-linear at cutoff with triangular or uniform kernel. Ad-hoc or user-supplied bandwidth.Limited
RD Robust (CCT bias-corrected)Sharp regression-discontinuity treatment effect via local-polynomial (triangular-kernel) fits on each side of the cutoff within an MSE-optimal bandwidth, reporting the conventional point estimate, the Calonico-Cattaneo-Titiunik (2014) bias-corrected estimate, conventional and robust standard errors, confidence intervals, and effective sample sizes.Full
Propensity Score Matching (PSM)Logit/probit propensity score; 1:1/1:k matching with replacement or caliper.Full
Extended Regression Model — Linear (ERM-Linear)Combines endogeneity (IV-2SLS), sample selection (Heckman), treatment effects in one model.Full
Extended Regression Model — Probit (ERM-Probit)Endogeneity (control function) + selection (bivariate-probit IMR) for binary outcome.Limited
Extended Regression Model — Ordered (ERM-Ordered)Ordered probit with sample selection correction via IMR augmentation.Limited
Extended Regression Model — Interval (ERM-Interval)Outcome observed as bounds [lower, upper]; handles left/right/interval censoring via MLE.Full
ERM DiagnosticsSpecification tests for endogeneity, selection bias, and treatment effect significance within ERM.Full
LOESS RegressionLocal polynomial regression; non-parametric smoothing via weighted least squares.Full
Gamma GLM (severity)Gamma generalized linear model with log link for strictly-positive responses (e.g. claim severity/cost); supports an exposure/offset term. Dispersion via Pearson, matching R glm(family=Gamma(link="log")).Full
Tweedie GLM (pure premium)Tweedie compound Poisson-Gamma GLM (log link) for non-negative data with a mass at zero — single-model pure-premium pricing; configurable variance power p in (1,2).Full

Post-Estimation & Diagnostics 47

Statios provides 31 post-estimation and regression diagnostic procedures spanning heteroskedasticity tests, serial correlation diagnostics, influence measures, marginal effects, contrasts, and model comparison tools.

ProcedureWhat it doesStatus
Normality tests (comprehensive suite)Shapiro-Wilk, Kolmogorov-Smirnov, D'Agostino-Pearson, Anderson-Darling; includes skewness/kurtosis and Q-Q plot data.Full
Homogeneity of variance testLevene and Bartlett tests for equality of variances across groups; concludes on Levene (robust).Full
Outlier detection (multiple methods)IQR, Z-score (|z|>3), and modified Z-score (MAD-based) methods with index identification.Full
Multicollinearity checkVariance Inflation Factor (VIF) analysis with concern classification (low/moderate/high).Full
Automatic test suggestionRecommends appropriate statistical tests based on data characteristics.Full
Breusch-Pagan test for heteroskedasticityLM and F variants test if residual variance depends on predictors; requires n > k+2.Full
White test for heteroskedasticityGeneral heteroskedasticity test not assuming specific form; more flexible than Breusch-Pagan.Full
Goldfeld-Quandt heteroskedasticity testSorts by first predictor, splits at the midpoint, drops a block of observations starting at the midpoint, and F-tests the lower vs upper group residual variances.Full
Breusch-Godfrey serial correlation testLM and F tests for autocorrelation up to specified lags in OLS residuals.Full
Durbin-Watson autocorrelation statisticDW near 2 indicates no autocorrelation; <1.5 positive, >2.5 negative autocorrelation.Full
Ljung-Box autocorrelation testTests joint significance of autocorrelations at specified lags; generic for series/residuals.Full
Heteroskedasticity-robust LM testWooldridge robust LM test that a set of regressors are jointly zero, valid under heteroskedasticity of unknown form (n − SSR auxiliary regression on residualized products).Full
AR(1) serial-correlation testWooldridge's t-test for first-order serial correlation: regress OLS residuals on their own lag (plus the regressors) and test ρ = 0.Full
Durbin's h TestDurbin's h test for first-order serial correlation in OLS models that include a lagged dependent variable, where the Durbin-Watson statistic is biased toward 2.Full
Moulton Factor (clustered SE inflation)Moulton (1986) standard-error inflation factor: estimates the residual intra-class correlation across groups and reports naive OLS vs Moulton-adjusted standard errors.Full
Vuong Test (non-nested)Vuong's (1989) likelihood-ratio test for two non-nested models (OLS/logit/Poisson), comparing per-observation log-likelihoods to decide which specification is closer to the true data-generating process, with AIC/BIC complexity-corrected variants.Full
Cox Test (non-nested)Pesaran-Cox test discriminating between two non-nested linear regression models via the difference in log-likelihoods adjusted by its expectation under each model, reported in both directions as standard-normal statistics with a preferred-model verdict.Full
Encompassing Test (non-nested)Mizon-Richard (1986) encompassing test that asks, by artificial nesting, whether one non-nested linear model explains everything its rival adds: each direction adds the rival's non-overlapping regressors to the base model and F-tests their joint significance (insignificant => the base model encompasses the rival), returning F/df/p, per-direction verdicts, and a preferred model.Full
J Test (non-nested models)Davidson-MacKinnon J test that discriminates between two non-nested linear specifications by adding each model's OLS fitted values to the other and testing their t-statistic in both directions, returning per-direction rejection verdicts and a preferred model.Full
Information-Matrix Test (White)White's (1982) omnibus information-matrix specification test for OLS: IM = n*R^2 of the auxiliary regression of squared residuals on the unique squares and cross-products of the regressors (chi-square), detecting general misspecification — wrong functional form, heteroskedasticity, or non-normality — not just heteroskedasticity.Full
Harvey Multiplicative HeteroskedasticityHarvey's (1976) LM test for multiplicative heteroskedasticity, regressing the log of squared OLS residuals on chosen variables to estimate the variance model Var(e)=exp(z'alpha) and test whether the variance depends on them.Full
LM / Score Test (omitted regressors)Classic Lagrange-multiplier (score) test for omitted regressors: regresses the residuals of the restricted OLS (dependent on kept) on the full regressor set and reports LM = n*R-squared of that auxiliary regression as a chi-square(q) test that the tested coefficients are jointly zero, with the equivalent F-form and a reject/not verdict.Full
Multivariate Normality (Doornik-Hansen)Tests whether a set of variables (or VAR residuals) is jointly multivariate normal via Mardia's multivariate skewness and kurtosis statistics and the Doornik-Hansen omnibus chi2(2p) test, returning each statistic with its df and p-value, per-variable skew/kurtosis, and an overall normal/non-normal verdict.Full
Ramsey RESET specification testTests for functional form misspecification by adding polynomial terms of fitted values.Full
Hosmer-Lemeshow goodness-of-fit testGroups logit predictions by deciles, compares observed vs expected event counts via chi-square.Full
Variance Inflation Factor (VIF) analysisVIF per predictor with concern thresholds (VIF>5 moderate, >10 high); includes mean VIF.Full
Cook's distance influence measureIdentifies observations with disproportionate influence on regression; threshold default 4/n.Full
Leverage (hat values) analysisDetects high-leverage observations (unusual X values); threshold 2*(k+1)/n.Full
DFBETAS parameter-specific influenceMeasures change in each coefficient when one observation removed; threshold 2/sqrt(n).Full
Marginal effects (logit/probit)Average Marginal Effects (AME) or Marginal Effects at Means (MEM) for binary choice models.Full
Predicted probabilities with CILogit/probit predictions on observed data or at specified covariate values with 95% CI.Full
Post-hoc multiple comparisonsTukey, Bonferroni, Holm step-down, Scheffe, Games-Howell with p-value adjustments.Full
Effect sizes across designsCohen's d, Glass's delta, Hedges' g, eta-squared, partial eta-squared, omega-squared, Cohen's f.Full
Planned contrasts (pairwise/polynomial)Pairwise, polynomial (Legendre), Helmert, deviation contrasts with t-tests and CI.Full
Prediction intervals (OLS)Confidence and prediction intervals for fitted values; wider than CI due to residual variance.Full
Coefficient table (with VIF/robust SE)Formatted OLS results with standardized betas, VIF, HC3 robust SE option; significance stars.Full
Model comparison (R-sq, AIC, BIC)Side-by-side OLS models with delta AIC/BIC relative to best; RMSE and F-stat per model.Full
Marginal effects at specified valuesConditional marginal effects (OLS/logit/probit/Poisson) at user-supplied covariate values.Full
Margins contrast (difference in effects)Compare marginal effects of one predictor across two or more levels; delta-method SE.Full
Conditional marginal effects (interaction)Marginal effect of focal variable at 10th, 25th, 50th, 75th, 90th percentiles of moderator.Full
Nested contrasts (hierarchical)Two-sample Welch t-tests on specified pairs within a grouping variable.Full
Elasticities (full/partial/arc)dy/dx*x/y (full), dy/dx*x (partial), or arc elasticity across specified predictor values.Full
Semi-elasticities (dy/dx/y and dy/dx*x)Proportional change in y and linear change in y per unit change in x.Full
Predicted means over groups (Stata-style)Marginal cell means with 95% CI; pairwise comparisons across factor levels.Full
Adjusted predictions (holding covariates constant)Linear predictor at specified covariate values (mean/median/min/max); 95% CI included.Full
Probability profile (logit/probit curve)Binary outcome probability as function of one predictor; other predictors at mean.Full
Treatment effect contrasts (ATE/ATT/ATC)Average treatment effect via regression-based g-computation; balanced or flexible interaction.Limited

Time Series & Macroeconometrics 84

Comprehensive time series analysis, DSGE models, and advanced forecasting. Includes unit root tests, ARIMA/seasonal models, VAR/cointegration, impulse responses, nonlinear models, forecast evaluation, and mixed-frequency regression.

ProcedureWhat it doesStatus
Autocorrelation & Partial Autocorrelation (ACF / PACF)Sample ACF and PACF (Durbin-Levinson / Yule-Walker) with white-noise confidence bands and per-lag Ljung-Box Q statistics, for ARMA order identification.Full
HAC Long-Run VarianceEstimates the heteroskedasticity- and autocorrelation-consistent (HAC) long-run variance S = gamma_0 + 2*sum k(j/L)*gamma_j of a time series using the Bartlett (Newey-West) or Quadratic-Spectral (Andrews) kernel, with the Newey-West automatic plug-in bandwidth, reporting the variance ratio S/gamma_0 and the HAC standard error of the mean.Full
Structural VECM (long-run restrictions)Fits a Johansen VECM at a given cointegration rank and identifies structural shocks via long-run (Granger-representation) restrictions, reporting the reduced-rank long-run impact matrix, the lower-triangular structural long-run matrix (K-r permanent / r transitory shocks), and few-horizon structural impulse responses.Full
Time-Varying-Parameter VAREstimates a VAR whose coefficients drift over time as random walks, cast as a linear-Gaussian state-space model and recovered by the Kalman filter and Rauch-Tung-Striebel smoother, returning smoothed coefficient paths (start/end/mean), state-innovation variances, an LR test of constant coefficients, and the log-likelihood.Full
Markov-Switching VARRegime-switching vector autoregression (Hamilton) whose intercepts, dynamics, and innovation covariance switch between latent Markov regimes, estimated by EM (Hamilton filter + Kim smoother) returning per-regime VAR coefficients, the regime transition matrix, smoothed regime probabilities, and log-likelihood.Full
Threshold VAR (TVAR)Estimates a two-regime multivariate threshold VAR by grid-searching the threshold that splits the sample on a transition variable's lag, fits a separate VAR in each regime, and tests for threshold nonlinearity with a bootstrap-calibrated sup-LR statistic.Full
Quantile VARVector autoregression estimated at a chosen conditional quantile via equation-by-equation quantile regression of each variable on the full lagged vector, returning per-lag tau-quantile coefficient matrices alongside the mean (OLS) VAR for comparison, to capture tail/regime-dependent dynamics.Full
Local Projections (Jorda IRF)Estimates Jorda (2005) local-projection impulse responses by running a separate OLS regression of y_{t+h} on the shock at t (plus controls) at each horizon with Newey-West HAC standard errors, yielding an impulse-response function with confidence bands that is robust to dynamic misspecification relative to VAR IRFs.Full
Unobserved Components (structural TS)Harvey structural time-series model: decomposes a univariate series into stochastic level, optional local-linear-trend slope, optional stochastic seasonal, and irregular via Kalman-filter maximum likelihood, reporting component variances, smoothed series, log-likelihood/AIC, and a short forecast.Full
Vector Moving-Average (VMA)Fits a multivariate VMA(q) model y_t = mu + eps_t + Theta_1 eps_{t-1} + ... + Theta_q eps_{t-q} by conditional Gaussian maximum likelihood, returning the MA coefficient matrices, innovation covariance, finite (zero-beyond-lag-q) impulse responses, log-likelihood and information criteria.Full
VARMA ModelEstimates a Vector ARMA(p,q) for a small multivariate time series by conditional maximum likelihood (via VARMAX), reporting the AR and MA coefficient matrices, residual covariance, log-likelihood and AIC/BIC.Full
Intervention Analysis (interrupted TS)Box-Tiao interrupted-time-series analysis fitting an AR(p) model with a deterministic step (permanent level shift) or pulse (transient shock) intervention dummy at a known date, reporting the estimated intervention effect with its t-test, the long-run steady-state level shift, pre/post means, and a plain-language interpretation.Full
Subset / Restricted VAREstimates a vector autoregression then imposes data-driven zero restrictions on coefficients whose full-VAR |t| falls below a threshold, re-estimating the sparser system by restricted EGLS and reporting retained coefficients, the number of restrictions, and the SC/HQ/AIC improvement over the full VAR.Full
Periodic VAR (seasonal coefficients)Fits a periodic vector autoregression / periodic autoregression (PAR) in which the intercepts and lag coefficient matrices vary deterministically by season, returns the per-season coefficient matrices, and tests the constant-coefficient (standard-VAR) restriction via per-equation F tests and a system likelihood-ratio test of periodicity.Full
Threshold Cointegration (Enders-Granger)Tests for threshold cointegration with asymmetric error correction by fitting Enders-Granger TAR and momentum-TAR models to the cointegrating-residual changes, reporting the two regime adjustment speeds, the Phi no-cointegration statistic, and an F-test of symmetric vs. asymmetric adjustment.Full
Generalized IRF (Koop-Pesaran-Potter)Computes ordering-invariant generalized impulse-response functions for a VAR (Koop-Pesaran-Potter 1996; Pesaran-Shin 1998), tracing each variable's response to a shock in the chosen variable without imposing a recursive Cholesky ordering.Full
LSTAR (Smooth-Transition AR)Fits a logistic smooth-transition autoregression by nonlinear least squares, estimating two linear AR regimes plus the smoothness (gamma) and threshold (c) of a logistic transition over a lagged level, with the Luukkonen-Saikkonen-Terasvirta LM test of linearity against smooth-transition nonlinearity.Full
Spectral Granger Causality (Geweke)Decomposes Granger causality from one series to another by frequency using Geweke's (1982) spectral measure of linear feedback, reporting causality at low / business-cycle / high frequencies, the dominant frequency, and the frequency-averaged measure that matches the time-domain Granger test by Geweke's integral identity.Full
Tsay Nonlinearity TestTsay (1986) F-test for threshold/quadratic nonlinearity in an AR(p) model: augments the linear AR with the squares and pairwise cross-products of the lagged values and tests their joint significance, returning the F statistic, degrees of freedom, p-value, and a linear/nonlinear verdict.Full
BDS Test (nonlinearity)Brock-Dechert-Scheinkman-LeBaron test for serial independence: computes the correlation-integral BDS statistic at embedding dimensions 2..max_dim and returns per-dimension N(0,1) statistics with p-values plus an overall iid / non-iid verdict, detecting both linear and nonlinear dependence (e.g. in model residuals).Full
Kalman Smoother (state-space)Fits a local-level (or local-linear-trend) unobserved-components state-space model by maximum likelihood and returns both the real-time Kalman-filtered and the full-sample RTS-smoothed level (and slope), with estimated irregular/level/slope variances, log-likelihood and smoothed-state standard errors.Full
Time-Varying-Parameter RegressionEstimates regression coefficients that evolve over time as random walks via a linear-Gaussian state-space model fit by the Kalman filter/smoother, returning the smoothed time-varying coefficient paths, signal/state variances, and log-likelihood to detect structural change.Full
IGARCH (Integrated GARCH)Integrated GARCH(p,q) estimated by maximum likelihood with the variance persistence (sum of ARCH+GARCH coefficients) constrained to exactly 1, so volatility shocks are permanent rather than mean-reverting.Full
GARCH-in-Mean (ARCH-M)Estimates a GARCH-in-mean (ARCH-M) model in which the conditional volatility (std. dev.) or conditional variance enters the mean equation as a risk-premium term, reporting the lambda risk premium with significance alongside the omega/alpha/beta variance dynamics and persistence.Full
FIGARCH (long-memory volatility)Estimates a fractionally-integrated GARCH(p,d,q) by ML over a truncated ARCH-infinity expansion, recovering the long-memory parameter d in (0,1) that places volatility persistence between the GARCH (d=0) and IGARCH (d=1) extremes, with conditional-volatility summary, log-likelihood, AIC/BIC and a long-memory verdict.Full
GARCH Portmanteau DiagnosticsFits a GARCH(1,1) by Gaussian QMLE, forms standardized residuals, and tests them for remaining structure via Ljung-Box Q on z_t and z_t^2 and Engle ARCH-LM on z_t^2 at every lag, returning the variance parameters, per-lag statistics with p-values, and a pass/fail adequacy verdict for the volatility specification.Full
Realized GARCHEstimates the Realized GARCH(1,1) model of Hansen, Huang & Shek (2012), jointly fitting the variance recursion h_t = omega + beta*h_{t-1} + gamma*x_{t-1} and the measurement equation x_t = xi + phi*h_t + tau(z_t) + u_t by Gaussian QMLE, reporting omega/beta/gamma, measurement parameters, persistence (beta+gamma*phi), log-likelihood, and the fitted conditional-volatility path (building a realized proxy from squared returns when no realized measure is supplied).Full
CCC Multivariate GARCHBollerslev (1990) Constant Conditional Correlation multivariate GARCH: fits a univariate GARCH(1,1) to each series, estimates the constant correlation matrix R from the standardized residuals, and forms conditional covariances H_t = D_t R D_t that are positive-definite by construction.Full
HAR-RV (realized volatility)Corsi (2009) Heterogeneous Autoregressive model of Realized Volatility: regresses next-period volatility on the daily (1-day), weekly (5-day) and monthly (22-day) trailing averages of a realized-volatility proxy, reporting component coefficients/SE/t/p, a joint significance F-test, R-squared and a one-step-ahead forecast.Full
BEKK Multivariate GARCHFits a diagonal BEKK(1,1) multivariate GARCH (Engle & Kroner 1995) by Gaussian ML to 2-3 return series, estimating the diagonal A/B ARCH/GARCH parameters, implied per-series persistence, the time-varying conditional covariance/correlation path (guaranteed positive-definite at every t), the mean dynamic correlation, and the log-likelihood.Full
Stochastic VolatilityFits a stochastic-volatility model whose latent log-variance follows its own AR(1) state, estimated by Harvey-Ruiz-Shephard quasi-ML via the Kalman filter on log-squared returns, returning the persistence (phi), vol-of-vol (sigma_eta), long-run variance level (mu), and the Kalman-smoothed conditional-volatility path.Full
Causality-in-Variance (Cheung-Ng)Cheung-Ng (1996) test for volatility spillover between two series: GARCH-filters each series, then uses the n-scaled sum of squared cross-correlations of the squared standardized residuals to test, in each direction, whether past volatility of one series predicts the other's future volatility.Full
FMOLS Cointegrating VectorPhillips-Hansen Fully-Modified OLS estimates a single cointegrating vector among I(1) series, correcting OLS for serial correlation and regressor endogeneity via a Bartlett long-run covariance to yield an asymptotically efficient, unbiased long-run coefficient table with valid standard errors.Full
DOLS Cointegrating VectorStock-Watson (1993) single-equation Dynamic OLS estimate of a cointegrating vector, regressing the I(1) dependent on the I(1) regressors plus leads and lags of their first differences, with HAC (Newey-West) standard errors on the asymptotically efficient long-run coefficients.Full
Markov Chain (transition/stationary)Estimates a first-order discrete Markov chain from a state sequence by maximum-likelihood transition counting, returning the row-normalized transition matrix, the stationary (invariant) distribution solving pi = pi P, the n-step transition matrix P^n, and Kac mean recurrence / mean first-passage times.Full
Phillips-Ouliaris Cointegration TestResidual-based test of the no-cointegration null among I(1) series: runs the cointegrating OLS, then applies the Phillips-Perron Z-alpha and Z-t unit-root statistics to the residuals (Newey-West long-run-variance corrected) against MacKinnon response-surface cointegration critical values to deliver Z-alpha, Z-t, an approximate p-value, and a cointegration verdict.Full
Blanchard-Quah SVARIdentifies permanent and transitory structural shocks in a bivariate (or n-variable) VAR via Blanchard-Quah long-run restrictions, returning the lower-triangular long-run impact matrix, the contemporaneous impact matrix, orthonormal structural impulse responses, and the forecast-error variance decomposition.Full
VAR Residual Diagnostics (multivariate)Fit a VAR(p) and run a battery of multivariate residual checks - the Hosking (1980) portmanteau/Ljung-Box test for serial correlation, a multivariate ARCH-LM serial-correlation test, and the Doornik-Hansen omnibus normality test - returning each statistic, df, p-value, and an overall model-adequacy verdict.Full
VAR Lag-Order Selection (IC table + LR)Fits a VAR(p) for every lag p=0..maxlags on a common fixed sample and reports the AIC, Hannan-Quinn, Schwarz/BIC and FPE information criteria plus a sequential likelihood-ratio test, returning each criterion's argmin and the LR-selected lag.Full
ARCH-LM Test (ARCH effects)Engle's LM test for conditional heteroskedasticity (volatility clustering): n·R² auxiliary regression of the squared series on its own lags. Run before fitting a GARCH model.Full
Time-trend regression (detrending)Regress a series on a polynomial deterministic time trend; reports trend coefficients, R², overall trend significance, and detrended-residual summary (trend- vs difference-stationary).Full
Spurious-regression diagnosticRegress y on x and flag the classic spurious-regression signature (high R², very low Durbin-Watson, non-stationary residuals via Engle-Granger ADF) for integrated, non-cointegrated series.Full
Seasonal ARIMA (SARIMA)Fit seasonal ARIMA with forecasting, including Ljung-Box and Jarque-Bera diagnostics.Full
Holt-Winters Exponential SmoothingExponential smoothing with additive/multiplicative seasonal component and trend.Full
Error-Trend-Seasonal (ETS) modelAutomated error-trend-seasonal decomposition with AIC/BIC selection.Full
Unobserved Components / State SpaceKalman filter state-space model with level, trend, and seasonal components.Full
Structural VAR (SVAR)Structural VAR with estimated A and B matrices for shock identification.Full
Impulse Response FunctionsOrthogonalized and non-orthogonalized IRF from VAR with cumulative responses.Full
Forecast Error Variance DecompositionVariance decomposition of VAR forecasts by shock source over horizons.Full
Markov Switching AutoregressionRegime-switching model with time-varying transition probabilities and smoothed probabilities.Full
Threshold Autoregression (TAR)Nonlinear autoregression with endogenous threshold detection and linearity test.Full
Dynamic Conditional Correlation GARCHDCC-GARCH model for time-varying correlations (Engle 2002) via quasi-MLE.Full
Zivot-Andrews unit root testUnit root test with one endogenous structural break in level or trend.Full
DF-GLS unit root testElliott-Rothenberg-Stock feasible GLS-detrended unit root test.Full
ERS Point Optimal unit root testEfficient detrending with GLS and point-optimal deviations under the null.Full
Change-in-mean structural break detectionDetects shifts in the MEAN of one series via dynamic-programming least-squares (L2) segmentation, with the number of breaks chosen by BIC; reports segment means and variances.Full
Quandt-Andrews Break Test (sup-Wald)Tests for a single structural break at an unknown date in a linear regression by computing the Chow/Wald F for a full-coefficient break at every trimmed candidate point and reporting the sup-F (QLR), average-F and exp-F (Andrews-Ploberger) statistics, the estimated break fraction, Andrews (1993) asymptotic critical values, and a Hansen (1997) approximate p-value.Full
CUSUM and CUSUM-squared stability testRecursive residual stability test for parameter constancy over time.Full
Particle Filter (SMC state-space)Bootstrap particle filter (sequential Monte Carlo) for a univariate local-level state-space model: estimates the process/observation variances by maximum likelihood on the exact Kalman likelihood, then runs an SIR filter with systematic resampling to return the filtered latent-state mean, a 90% credible band, the one-step predictive log-likelihood, and the effective sample size over time, benchmarked against the exact Kalman filter.Full
Koyck Distributed LagEstimates a Koyck (geometric) infinite distributed-lag model y_t = alpha + beta*x_t + lambda*y_{t-1} via OLS on the Koyck transform, reporting the impact multiplier, geometric decay rate, long-run multiplier beta/(1-lambda), mean/median lag, the implied lag-weight profile, and a Durbin's-h serial-correlation diagnostic.Full
Toda-Yamamoto Granger causalityGranger causality test in levels with integration-order adjustment; Wald test on VAR lags.Full
Nonlinear ARDL (NARDL)Autoregressive distributed lag with asymmetric positive/negative decomposition; symmetry test.Full
Pesaran-Shin-Smith ARDL bounds testCointegration test via F-test on lagged levels in error-correction form.Full
Diebold-Mariano forecast comparisonTest comparing forecast accuracy of two models (MSE or MAE) with Newey-West variance.Full
Forecast combinationCombine forecasts via equal weight, inverse MSE, OLS, or Bates-Granger method.Full
Diffusion-index direct forecast (user-supplied factors)Direct multi-step diffusion-index forecast y_{t+h} = a + sum g_p y_{t-p} + b'F_t using factor columns you supply (e.g. PCA component scores from analysis.pca), with an out-of-sample backtest against an AR benchmark.Limited
Dynamic Factor Model (EM/Kalman)Estimate a latent dynamic factor model on multiple standardized time series by EM with the Kalman filter/RTS smoother (Doz-Giannone-Reichlin / Watson-Engle), extracting common AR factors and reporting loadings, the smoothed factor path, factor AR coefficients, variance shares, log-likelihood and iteration count.Full
Spectral analysisPeriodogram or Welch PSD with dominant frequency and period identification.Full
Hamilton Filter (trend/cycle)Hamilton (2018) regression-based business-cycle filter that decomposes a series into trend (the h-step-ahead forecast fitted from a constant and p lagged levels) and cycle (the forecast residual), reporting coefficients, R^2 and the cycle standard deviation as a sound, spurious-dynamics-free replacement for the Hodrick-Prescott filter.Full
Spectral Coherence (cross-spectrum)Estimates the squared coherency and phase spectrum between two time series via a smoothed Welch/averaged-periodogram cross-spectrum with a Hann taper, reporting coherence and phase by frequency, the peak-coherence frequency, and an approximate significance threshold.Full
Unrestricted MIDAS regressionMixed-frequency regression: unrestricted lags of high-frequency predictor on low-frequency outcome.Limited
Almon polynomial MIDASMIDAS with polynomial lag weighting via NLS; information criterion comparison across lag lengths.Limited
Beta polynomial MIDASMIDAS with beta lag structure for smooth, parsimonious high-frequency aggregation.Limited
Exponential Almon MIDASMIDAS using exponential-Almon lag polynomial for flexible decay pattern.Limited
MIDAS forecastForecast using fitted MIDAS model; extends from low-freq observations with high-freq predictors.Full
MIDAS nowcastingReal-time estimate of current low-frequency value using available high-frequency data.Limited
MIDAS model comparisonCompare U-MIDAS, Almon, Beta, and Exponential MIDAS across lag range via information criteria.Limited
Real Business Cycle modelCanonical RBC with technology shock; analytical steady state, IRF, variance decomposition.Limited
New Keynesian 3-equation modelIS–Phillips–Taylor system; parameterizable demand, supply, monetary shocks with closed-form IRF.Limited
Linear rational expectations solverKlein-Sims QZ method for general linear RE systems; determines Blanchard-Kahn determinacy.Full
DSGE simulationForward simulation of solved DSGE policy matrices under specified or random shock sequence.Full
DSGE impulse responseImpulse response to single shock in solved DSGE model; orthogonal identification.Full
DSGE variance decompositionForecast error variance share by shock at specified horizons in DSGE.Full
DSGE calibrationHeuristic moment-matching calibration for RBC and NK; suggests parameters given targets.Limited

Panel Data & Causal Inference 51

Statios supports canonical panel data estimators (FE, RE, Hausman test, 2SLS-IV), advanced dynamic panel methods (Arellano-Bond, System GMM), and modern causal inference designs (DiD, RDD, synthetic control, staggered DiD, IPW, mediation analysis).

ProcedureWhat it doesStatus
Fixed Effects Panel RegressionWithin-transformation OLS eliminating time-invariant unobserved heterogeneity. Optional two-way (entity + time) fixed effects.Full
Random Effects Panel RegressionGLS panel estimator assuming random effects uncorrelated with regressors.Full
Hausman Specification TestTests FE vs RE using Moore-Penrose generalized inverse when covariance difference non-positive semidefinite (finite-sample robust).Full
Two-Stage Least Squares (2SLS) IVInstrumental variables regression with heteroskedasticity-robust first-stage F-test and weak-instrument diagnostics.Full
Arellano-Bond Difference GMMDynamic panel estimator using lagged levels as instruments for differenced equations; includes Sargan and AR(1)/AR(2) tests.Full
Blundell-Bond System GMMTwo-step efficient GMM stacking level and difference equations; Hansen J test and AR tests.Full
Hausman-Taylor EstimatorEstimates time-varying exogenous, time-invariant exogenous, and endogenous coefficients using within and between equations.Full
Between EstimatorOLS on entity-level means; removes within-group variation.Full
First-Difference EstimatorEliminates time-invariant heterogeneity via differencing; OLS on changes.Full
Panel-Corrected Standard Errors (Beck-Katz PCSE)Kernel-based covariance estimator robust to contemporaneous and lagged cross-sectional dependence.Full
Driscoll-Kraay Standard ErrorsRobust to both cross-sectional and temporal dependence via time clustering.Full
Panel Cointegration TestsPedroni, Kao, or Fisher method tests for long-run equilibrium relationships in panel data.Full
Panel Granger Causality (Fisher Combined)Granger test per entity combined via Fisher's method across units.Full
Cross-Sectional Dependence TestPesaran CD, Frees, or Friedman test for contemporaneous cross-unit correlation.Full
Panel ARDL (Mean Group / Dynamic Fixed Effects)Panel autoregressive distributed lag models: Mean Group (averages unrestricted per-entity long-run estimates) and Dynamic Fixed Effects (pooled with entity intercepts). Reports short-run, long-run, and ECM terms.Full
Mundlak (Correlated Random Effects)RE model augmented with entity means; joint test on means equivalent to Hausman test; robust to cluster heteroskedasticity.Full
Panel RE Test (Breusch-Pagan LM)Breusch-Pagan Lagrange-multiplier test for entity random effects in a panel: from pooled-OLS residuals it forms the two-sided BP-LM (chi-squared, 1 df) and the more powerful one-sided Honda statistic (normal) from grouped within-entity residual sums, handles balanced and unbalanced panels, and verdicts whether pooled OLS is adequate or a random/fixed-effects estimator is required.Full
Panel Serial-Correlation Test (Wooldridge)Wooldridge (2002) F-test for first-order serial correlation in the idiosyncratic errors of a linear panel model -- first-differences the equation, regresses the differenced residuals on their own lag and tests whether the slope equals the no-serial-correlation value of -0.5 (cluster-robust, as Stata xtserial), and also reports the Baltagi-Wu LBI.Full
Anderson-Hsiao Dynamic PanelAnderson-Hsiao IV estimator for dynamic panels: first-differences out the fixed effect, then instruments the differenced lagged dependent with its t-2 level via 2SLS, recovering a consistent persistence coefficient where pooled OLS and within are biased.Full
FOD-GMM Dynamic Panel (Helmert)Arellano-Bover (1995) forward-orthogonal-deviations (Helmert) GMM for dynamic panels: removes fixed effects via the FOD transform (which keeps the idiosyncratic errors serially uncorrelated, so the one-step weight is simply (Z'Z)^-1), instruments the transformed lagged dependent with the expanding set of lagged levels, and reports coefficient estimates (coef/se/z/p) with entity-clustered robust SEs and a Hansen J over-identification test - markedly less biased than the within estimator for short T.Full
Bias-Corrected FE (dynamic panel)Within (LSDV) fixed-effects estimation of a dynamic panel with a lagged dependent variable, then the analytical Nickell (1981) / Kiviet (1995) correction of the finite-T within (Nickell) bias subtracted from the autoregressive coefficient, returning the uncorrected and bias-corrected persistence plus the bias estimate.Full
Parks-Kmenta FGLS (long panel)Feasible GLS for long balanced panels (T > N) that simultaneously corrects for panel heteroskedasticity, contemporaneous cross-sectional correlation, and panel-specific AR(1) serial correlation, returning efficient slope estimates, per-entity AR(1) coefficients, and Parks-Kmenta standard errors (with the standard over-optimism caveat).Full
Common Correlated Effects (CCE)Pesaran (2006) CCE estimator for heterogeneous panels with cross-sectional dependence: augments each unit's regression with cross-sectional averages of the dependent and regressors to proxy and absorb the unobserved common factors, then reports the Mean-Group (CCEMG) or Pooled (CCEP) slopes, which remain consistent where plain fixed effects is biased by the factor.Full
Panel Quantile Regression (Canay)Estimates fixed-effects panel quantile regression via Canay's (2011) two-step estimator -- removing each entity's location-shift fixed effect from a within mean regression, then running ordinary quantile regression at tau on the de-fixed data -- yielding slope effects on the conditional tau-quantile that are robust to entity fixed effects (unlike biased pooled quantile regression).Full
Error Components + AR(1) (Baltagi-Li)One-way random-effects panel whose idiosyncratic error follows an AR(1) process: estimates the autocorrelation parameter rho, applies the Baltagi-Li Prais-Winsten transform and Sherman-Morrison FGLS, reports the variance components, the FGLS coefficient table, and a test of rho=0 (no serial correlation).Full
Pooled Mean Group (PMG)Estimates a dynamic heterogeneous panel ARDL error-correction model with common (pooled) long-run coefficients and unit-specific short-run dynamics and adjustment speeds via the Pesaran-Shin-Smith concentrated-likelihood iteration, reporting pooled long-run coef/se/z/p, the average error-correction speed, per-unit speeds, and a Hausman MG-vs-PMG poolability test.Full
MINQUE/MIVQUE Variance ComponentsEstimates the between-entity and idiosyncratic variance components of a one-way error-components panel by Rao's non-iterative Minimum-Norm/Minimum-Variance Quadratic Unbiased estimator (handles unbalanced panels cleanly, unlike ANOVA), reports the implied intraclass correlation, and forms the FGLS slope estimates.Full
Random-Trend PanelFixed-effects panel model that gives every entity its own intercept AND its own linear time trend (swept out via a double-within detrending transform), recovering the common predictor slopes purged of the unit-specific trends that would bias a plain one-way FE estimator.Full
BLUP of Random EffectsPredicts each entity's unobserved random effect after a one-way random-effects (GLS) panel fit by shrinking its mean composite residual toward zero with the Henderson BLUP factor theta_i = sigma_u^2/(sigma_u^2 + sigma_e^2/T_i), reporting per-entity BLUPs, the shrinkage factor, and the variance components.Full
Two-Way Random EffectsTwo-way error-components GLS panel model with random entity AND time effects, estimating the entity, time, and idiosyncratic variance components (Wallace-Hussain) and reporting the Fuller-Battese theta transforms and FGLS coefficients.Full
Panel Unit Root (IPS / LLC)First-generation panel unit-root tests on a long panel: the Im-Pesaran-Shin W-bar statistic (standardised average of per-entity augmented Dickey-Fuller t-stats, allowing heterogeneous AR roots) plus the Levin-Lin-Chu pooled t* (common-rho, bias-adjusted), each returned with its N(0,1) p-value, the per-entity ADF t-stats and a stationary/unit-root verdict.Full
Panel Cross-Sectional Dependence (BP-LM)Tests panel errors for cross-sectional dependence from within-model residuals: the Breusch-Pagan LM (sum of T-weighted squared pairwise residual correlations ~ chi2(N(N-1)/2)), the Pesaran (2004) scaled LM, the Pesaran-Ullah-Yamagata (2008) bias-corrected scaled LM, and the reference Pesaran CD statistic, each with p-values and a verdict.Full
CIPS Panel Unit Root (2nd-gen)Pesaran (2007) cross-sectionally augmented IPS panel unit-root test, robust to cross-sectional dependence from a common factor: it averages per-entity CADF t-stats (each ADF regression augmented with the cross-section means of the lagged level and the differences) and compares the CIPS statistic against the non-standard Pesaran critical values.Full
Panel Poolability (Chow-F)Chow-type F test of panel poolability comparing restricted pooled-OLS SSR against the sum of unrestricted per-entity OLS SSRs to assess whether intercepts and slopes are homogeneous across entities, F~F((G-1)(k+1), N-G(k+1)).Full
Difference-in-Differences (DiD)2x2 DiD with cluster-robust or HC1 SEs; includes group means and specification.Full
Regression Discontinuity (Sharp RDD)Local polynomial with triangular or uniform kernel; ad-hoc 1.5*SD bandwidth (not MSE-optimal) or user-specified.Limited
Propensity Score Matching (PSM)Logistic propensity score with 1-NN matching on logit scale; caliper on logit scale (Austin 2011); naive paired SE (Abadie-Imbens not applied).Limited
Chow Structural Break TestTests equality of regression coefficients across two subsamples; F-statistic with df protection.Full
Synthetic Control MethodWeighted combination of donors minimizing pre-treatment MSE; RMSPE ratio and period-level treatment effects.Full
Event-study (dynamic two-way fixed effects)Conventional dynamic TWFE event study: two-way FE with relative-time dummies; pre-treatment trend test; reference period omitted.Full
Staggered DiD (Callaway & Sant'Anna 2021)Group-time ATT estimates with not-yet-treated comparison group; bootstrap SEs; dynamic, group, and calendar aggregations.Full
IV DiagnosticsFirst-stage F-tests, Cragg-Donald Wald F, Sargan over-identification, Durbin endogeneity test.Full
Inverse-Probability Weighting (IPW)Logistic propensity score weighting; ATE, ATT, or ATU estimands.Full
Doubly Robust EstimationOutcome model + propensity weighting; consistent if either model correct.Full
Local Average Treatment Effect (LATE)IV-based LATE for compliers only; asymptotic SE under 2SLS.Full
Classical Mediation AnalysisClassical Baron-Kenny path decomposition (direct/indirect effects) with a Sobel test and a percentile bootstrap CI for the indirect effect.Full
Sensitivity AnalysisE-value or Rosenbaum bounds for unobserved confounding robustness.Full
Lee BoundsNonparametric bounds under sample selection on unobservables.Full
Manski Treatment BoundsWorst-case partial-identification (Manski 1990) bounds on the average treatment effect of a binary treatment, with monotone-treatment-response and monotone-treatment-selection refinements that tighten the identified ATE interval.Full
Sharp Regression Kink Design (RKD)Sharp RKD: estimates the change in the outcome's slope at the threshold via piecewise-linear local regression (assumes a known unit kink in the assignment-to-policy map).Full
Bunching Estimator (normalized excess mass)Estimates excess mass at a policy threshold by fitting a counterfactual polynomial density excluding the bunching region; reports the normalized excess mass (bunching ratio).Full

Multivariate, Factor & Latent-Variable Analysis 35

Covers exploratory and confirmatory dimensionality reduction, clustering, reliability, latent structure discovery, and item response modeling — from classical PCA/factor analysis through modern IRT, SEM, and categorical methods.

ProcedureWhat it doesStatus
Principal Component AnalysisEigenvalue-based dimensionality reduction via correlation-matrix SVD with Kaiser criterion for retention.Full
Exploratory Factor AnalysisML factor extraction with Varimax/Promax rotation, KMO and Bartlett's sphericity test.Full
K-Means ClusteringStandardized iterative partition with silhouette score; sample-based for n > 10k.Full
Hierarchical Agglomerative ClusteringDistance-based clustering with Ward/complete/average linkage; limited to n ≤ 20k (O(n²) memory).Limited
Cronbach's AlphaInternal consistency reliability via item correlations with item-total and alpha-if-deleted.Full
McDonald's Omega (Total)ML single-factor CFA reliability coefficient without tau-equivalence assumption; requires k ≥ 3 items.Full
Split-Half ReliabilityOdd-even split with Spearman-Brown and Guttman corrections.Full
Parallel AnalysisHorn's method: compares observed eigenvalues against 95th percentile of random-data eigenvalues.Full
Multivariate Analysis of VarianceOne-way MANOVA with Wilks' lambda, Pillai trace, Hotelling-Lawley, Roy's root and F-approximations.Full
Linear Discriminant AnalysisFisher discriminant functions and linear classification functions with resubstitution accuracy.Full
Canonical Correlation AnalysisMaximum correlation between two variable sets via CCA with Rao's F-approximation for Wilks' lambda.Full
Confirmatory Factor AnalysisMeasurement model via semopy with factor loadings, residuals, and model fit (CFI, TLI, RMSEA, SRMR).Full
Structural Equation ModelingFull latent-variable SEM (measurement + structural paths) via semopy with standardized estimates.Full
Path AnalysisSingle-level directional relationships with path coefficients, R² per endogenous, and overall fit.Full
Rasch Model (1PL IRT)Joint MLE for dichotomous items; person abilities, item difficulties, infit/outfit, reliability.Full
2PL IRT ModelTwo-parameter logistic with discrimination + difficulty; uses girth if available, else manual L-BFGS.Full
3PL IRT ModelThree-parameter logistic (discrimination, difficulty, guessing) via girth or manual EM.Full
Graded Response Model (Samejima)Ordinal polytomous items with cumulative thresholds per item; category response curves and item info.Full
Partial Credit Model (Masters)Polytomous step parameters without item discrimination; flexible by-item category counts.Full
Rating Scale Model (Andrich)Polytomous with common thresholds across items and item-specific location parameters.Full
Nominal Response Model (Bock)Unordered categorical responses with per-category slopes and intercepts per item.Full
Differential Item Functioning AnalysisMantel-Haenszel, logistic, or Lord's chi-squared DIF with ETS delta classification.Full
Classical Test Theory AnalysisItem difficulty, corrected item-total correlation, and alpha-if-deleted per item.Full
IRT Fit Indices (Approximate)Limited-information chi-square, approximate RMSEA/CFI, Q3 local independence (NOT M2-based).Limited
Multidimensional ScalingMetric or non-metric embedding via Euclidean distances; stress and Shepard diagram.Full
Simple Correspondence AnalysisSVD on contingency-table chi-square distances for two categorical variables.Full
Multiple Correspondence AnalysisIndicator-matrix CA for k ≥ 2 categorical variables; uses prince or manual SVD.Full
Gaussian Finite Mixture Modelk-component Gaussian mixture with BIC/AIC selection, posteriors, and component weights.Full
Latent Class AnalysisEM-estimated discrete latent classes for binary/categorical indicators; entropy and model fit.Full
Latent Profile AnalysisDiagonal-covariance GMM for continuous indicators with entropy, classification table.Full
Categorical Regression (CATREG)ALS optimal scaling for nominal/ordinal predictors with regression fit on quantified variables.Full
Categorical PCA (CATPCA)Optimal scaling PCA with ALS alternation; preserves ordinal/nominal structure.Full
OVERALS (Non-linear CCA)Multi-set non-linear canonical correlation via ALS with set-wise quantifications.Full
Optimal Scaling TransformApplies nominal/ordinal/spline quantifications to variables without modeling.Full
Categorical CorrelationPearson correlations on optimally-scaled categorical variables.Full

Survival, Bayesian, Meta-Analysis & Resampling 68

Statistical methods for survival analysis, Bayesian inference, meta-analytic pooling, and resampling-based inference. Includes Kaplan-Meier curves, Cox proportional-hazards regression, parametric and competing-risks models, Bayesian linear/logistic/hierarchical regression with MCMC diagnostics, fixed- and random-effects meta-analysis, and bootstrap/jackknife/permutation methods.

ProcedureWhat it doesStatus
Kaplan-Meier Survival CurvesNon-parametric survival estimation with Greenwood pointwise 95% CI and log-rank/multivariate log-rank test.Full
Cox Proportional HazardsSemi-parametric hazard regression with Efron tie handling, PH assumption test (Grambsch-Therneau), and concordance index.Full
Mixture Cure ModelFits a Berkson-Gage mixture cure model by EM, separating a logistic incidence submodel for the long-term-survivor (cured) fraction from a Weibull AFT latency submodel for the susceptible subjects who can still experience the event.Full
Piecewise-Exponential SurvivalProportional-hazards survival model with a piecewise-constant baseline hazard, fitted by the Poisson / piecewise-exponential equivalence on person-interval exposure with covariate log-hazard-ratios; reduces to the exponential model with one interval.Full
Discrete-Time Hazard (cloglog)Fits a discrete-time (grouped-duration) proportional-hazards model by expanding subjects into person-period rows and estimating a complementary-log-log binomial GLM, yielding constant-across-time log-hazard-ratios that mirror the Cox model at low per-period hazard, plus a nonparametric or polynomial baseline-hazard summary.Full
Actuarial Life TableCutler-Ederer life table with conditional survival/death probabilities, cumulative hazard rate, and Greenwood SE.Full
Parametric Survival ModelsUnivariate or AFT parametric models (Weibull, exponential, lognormal, log-logistic, Gompertz) with parameter estimates.Full
Accelerated Failure TimeAFT regression with covariates, time ratio estimates, concordance index, and AIC.Full
Competing Risks (Fine-Gray)Subdistribution-hazards model with cumulative incidence functions per event type and subdistribution HR.Full
Shared Frailty Cox ModelCluster-robust Cox with gamma frailty variance estimate and LR test vs no-frailty baseline.Limited
Schoenfeld Residuals TestPH assumption test per covariate and globally using scaled Schoenfeld residuals with interpretation.Full
Nelson-Aalen Cumulative HazardNon-parametric cumulative hazard estimator with 95% pointwise confidence bands, optionally grouped.Full
Stratified Log-Rank TestLog-rank test with optional stratification, multivariate for 3+ groups, and pairwise comparisons.Full
Restricted Mean Survival TimeRMST up to user-specified tau with SE and 95% CI; difference test and CI for 2 groups.Full
Time-Varying Cox ModelCox regression with time-varying covariates (start-stop format), HR with CI and partial log-likelihood.Full
Cox Residuals for DiagnosticsFive residual types (martingale, deviance, Schoenfeld, scaled Schoenfeld, score) for model diagnostics.Full
Bayesian Linear RegressionMCMC-based (PyMC or emcee fallback) with weakly informative priors, posterior samples, PPC, and diagnostics (R-hat, ESS).Full
Bayesian Unit-Root AnalysisClosed-form Student-t posterior for the AR(1) root rho under a flat/Jeffreys prior, reporting the posterior mean/sd, a 95% credible interval, and the posterior probabilities P(rho>=1) (unit root/explosive) and P(rho<1) (stationary).Full
Bayesian VAR (Minnesota prior)Estimates a reduced-form VAR(p) with the Minnesota (Litterman) shrinkage prior imposed by dummy observations, reporting posterior-mean coefficient matrices that center the own first lag at 1 (levels) or 0 (stationary), shrink cross-variable lags harder than own lags by 1/lag, plus the shrinkage hyperparameters, per-equation in-sample fit, and a multi-step forecast.Full
Bayesian VECM (cointegration)Bayesian Vector Error Correction Model for cointegrated series: given a chosen cointegrating rank and lag order, places a weakly-informative Normal-inverse-Wishart prior on the short-run dynamics and adjustment coefficients and Gibbs-samples the posterior, reporting posterior means and 95% credible intervals for the speed-of-adjustment loadings (alpha), the normalised cointegrating vectors (beta), the short-run matrices, and error-correction evidence.Full
Bayesian CFA (one-factor)Fits a one-factor confirmatory measurement model x_ij = lambda_j * eta_i + eps_ij by a three-block Gibbs sampler (factor scores, loadings, residual variances), returning posterior loading summaries with 95% credible intervals, residual (uniqueness) variances, implied communalities, and an SRMR-based fit summary.Full
Bayesian IRT (2PL)Bayesian two-parameter logistic item-response model fitted by a Metropolis-within-Gibbs MCMC sampler, returning posterior item discriminations and difficulties (with 95% credible intervals) plus latent person-ability summaries on the N(0,1)-identified scale.Full
Bayesian Meta-AnalysisFits a random-effects meta-analysis (theta_i ~ N(mu, tau^2), y_i ~ N(theta_i, se_i^2)) by a direct Gibbs sampler with a weakly-informative half-Cauchy prior on the between-study sd, returning the posterior overall effect mu with credible interval, between-study heterogeneity tau, shrunken per-study estimates, and the posterior probability the effect is positive.Full
Conjugate Bayesian Regression (Normal-Gamma)Analytic conjugate Bayesian linear regression with a Normal-Gamma (Normal-inverse-Gamma) prior that returns closed-form multivariate-t posteriors for the coefficients and an inverse-Gamma posterior for the error variance, reducing exactly to OLS under a flat prior and shrinking slopes toward zero under a tight prior, with no MCMC.Full
Bayesian Quantile RegressionEstimates the conditional tau-quantile of an outcome via the asymmetric-Laplace likelihood, sampling the posterior of the quantile-regression coefficients with the Kozumi-Kobayashi (2011) location-scale-mixture Gibbs sampler and reporting posterior means, sds and 95% credible intervals.Full
Student-t (robust) RegressionOutlier-robust linear regression with iid Student-t(nu) errors fitted by an EM / IRLS scale-mixture ML algorithm, downweighting extreme residuals so the fit tracks the bulk of the data rather than the outliers, and reporting coefficients/SEs, the error scale, the (optionally ML-estimated) degrees of freedom, log-likelihood and per-observation robustness weights.Full
Metropolis-Hastings (Bayesian regression)Random-walk Metropolis-Hastings MCMC sampler for the joint posterior of regression coefficients and the error variance (beta, log sigma^2) in Bayesian linear regression, with a proposal auto-tuned toward the optimal ~0.234 acceptance rate, reporting posterior means/sds/95% credible intervals, the acceptance rate, and effective sample sizes.Full
Gibbs Sampler (Bayesian regression)Conjugate Gibbs sampler for Bayesian linear regression that alternates Normal draws of the coefficients given sigma^2 and inverse-Gamma draws of sigma^2 given the coefficients, returning posterior means/sds/95% credible intervals, effective sample size and a Geweke convergence diagnostic (acceptance is identically 1), with weak-prior posteriors that coincide with OLS.Full
Laplace Approximation (Bayesian)Approximates the posterior of a Bayesian linear or logistic regression by a Gaussian centered at the posterior mode (MAP) with inverse-Hessian covariance, reporting approximate posterior means/SDs/credible intervals, the Laplace approximation to the log marginal likelihood (model evidence), and a Tierney-Kadane fully-exponential refined posterior mean.Full
Importance Sampling (Bayesian)Bayesian posterior inference for a linear regression by importance sampling / sampling-importance-resampling: proposals drawn from a heavy-tailed multivariate-t centred at OLS (covariance inflated) under a weakly-informative conjugate Normal-inverse-Gamma prior are reweighted to give importance-weighted posterior means, 95% credible intervals for the coefficients and error variance, the weights' effective sample size, and a resampled (SIR) posterior draw.Full
Bayesian Poisson RegressionFits a log-linear Poisson count model by random-walk Metropolis-Hastings under a weak Normal prior, returning posterior means, standard deviations and 95% credible intervals for each coefficient, the incidence-rate ratios exp(coef), and the Metropolis acceptance rate (tuned toward the ~0.234 optimal scaling).Full
Bayesian TobitLeft-censored (Tobit) regression estimated by Gibbs sampling with data augmentation of the censored latent responses, returning posterior means and 95% credible intervals for the coefficients and error standard deviation plus the censored fraction.Full
Bayesian Probit (Albert-Chib)Fits a binary probit model by Gibbs sampling with Albert-Chib truncated-normal latent-utility data augmentation under a diffuse Normal prior, returning posterior means, SDs, 95% credible intervals, effective sample sizes, and the frequentist MLE as a reference.Full
Horseshoe Regression (sparse)Bayesian linear regression with a horseshoe shrinkage prior (Carvalho-Polson-Scott 2010) sampled by the fully-conjugate Makalic-Schmidt (2016) inverse-gamma scale-mixture Gibbs sampler, reporting posterior means and 95% credible intervals per coefficient, per-coefficient shrinkage weights kappa = 1/(1 + n*tau^2*lambda^2), and which coefficients are effectively selected (CI excludes 0) so that on sparse designs the null coefficients are shrunk hard toward 0 while genuine signals stay near their true values.Full
Spike-and-Slab Variable SelectionBayesian stochastic search variable selection (George-McCulloch SSVS) that places a two-component spike-at-zero / diffuse-slab prior on every regression coefficient and Gibbs-samples Bernoulli inclusion indicators to report each predictor's posterior inclusion probability (PIP), Bayesian-model-averaged coefficients, and the most probable visited models.Full
Bayesian LassoBayesian lasso regression (Park & Casella) places a Laplace prior on the coefficients and samples it with a scale-mixture-of-normals Gibbs sampler, returning shrunk posterior-mean coefficients with 95% credible intervals, the posterior of the shrinkage parameter lambda, and which predictors are effectively selected (CI excludes 0).Full
Adaptive LassoWeighted lasso (Zou 2006) whose per-coefficient penalties come from an initial OLS/ridge estimate, delivering oracle variable selection: truly-zero coefficients are driven to exactly 0 while genuine signals stay nearly unbiased, with the penalty chosen by reproducible cross-validation.Full
Savage-Dickey Bayes FactorComputes the Savage-Dickey density-ratio Bayes factor (BF01/BF10) for the sharp null H0: beta=0 on one regression coefficient by taking the ratio of the marginal posterior to the prior density at zero under a conjugate Normal-prior Bayesian regression, returning posterior mean/sd, the prior sd, and a Jeffreys-scale verdict.Full
Bayesian IC (DIC/WAIC/LOO)Bayesian predictive information criteria for a conjugate-Gibbs Bayesian linear regression, reporting DIC (Spiegelhalter), WAIC (Watanabe) and Pareto-smoothed importance-sampling LOO (Vehtari-Gelman-Gabry) on the deviance scale with their effective-parameter counts (p_DIC, p_WAIC, p_LOO) and Pareto-k diagnostics, for principled out-of-sample model comparison.Full
Bayesian Model AveragingEnumerates all 2^k linear-regression covariate subsets, scores each by BIC to obtain approximate posterior model probabilities, and reports per-predictor posterior inclusion probabilities (PIP), BMA posterior means/sds of the coefficients (averaged over models), and the top models.Full
Bayesian Logistic RegressionBernoulli likelihood with logit link, posterior odds ratios, in-sample accuracy, and MCMC diagnostics.Full
Bayesian Hierarchical ModelRandom-intercept multilevel model (PyMC, emcee, or OLS shrinkage fallback) with ICC, varying intercepts per group.Limited
MCMC DiagnosticsGelman-Rubin R-hat, ESS via autocorrelation, trace summaries with interpretations of convergence.Full
Bayes FactorModel comparison via BIC approximation with Jeffreys-scale interpretation (decisive/strong/moderate/anecdotal).Limited
Prior vs PosteriorVisualization comparison of prior and posterior distributions per parameter.Full
Posterior Predictive CheckBayesian p-value and visual comparison of observed vs simulated data from posterior predictive.Full
Credible IntervalHighest density interval (HDI) and equal-tailed interval (ETI) at 94% credibility.Full
Fixed-Effects Meta-AnalysisInverse-variance weighting, Cochran Q, I², per-study and pooled CIs, assumes homogeneous effects.Full
Random-Effects Meta-AnalysisDerSimonian-Laird tau² estimation, I², 95% prediction interval (Higgins-Thompson-Spiegelhalter), per-group weights.Full
Egger's Test for Publication BiasFunnel-plot asymmetry test via standardized-effect intercept regression, t-test on k-2 df.Full
Forest Plot DataPer-study effects with 95% CI, pooled estimate (fixed or random), I², and heterogeneity for visualization.Full
Bootstrap Standard ErrorBootstrap distribution of a statistic (mean/median/std/var/quantiles) with SE, bias, and summary.Full
Wild Bootstrap (OLS)Wild (and wild-cluster) bootstrap standard errors, percentile CIs, and p-values for OLS coefficients, robust to heteroskedasticity and the standard tool for few-cluster inference.Full
Percentile-t Bootstrap (coefficient)Studentized (bootstrap-t) confidence interval for a single OLS regression coefficient, built by pairs-resampling the pivot t*=(b*-b)/se* and inverting it via the Hall pivot-flip formula for second-order-accurate, skewness-corrected coverage, contrasted against the normal-theory b +/- 1.96 se interval.Full
Block Bootstrap (time series)Moving-block and circular-block bootstrap that resamples overlapping blocks of consecutive observations to deliver dependence-robust standard errors and percentile confidence intervals for the mean, median, variance, or lag-1 autocorrelation of a time series.Full
Parametric Bootstrap (OLS)Resamples OLS errors from the fitted homoskedastic-Gaussian model (y* = X*beta_hat + N(0, sigma^2)), refits each synthetic sample, and reports per-coefficient bootstrap SEs and percentile 95% CIs that converge to the classical OLS standard errors as replications grow.Full
Residual Bootstrap (OLS)Model-based (residual) bootstrap that holds the design and fitted values fixed, resamples OLS residuals with replacement to build synthetic responses, refits each, and reports per-coefficient bootstrap standard errors and percentile 95% confidence intervals without assuming any error distribution.Full
Subsampling InferencePolitis-Romano-Wolf subsampling: draws subsamples of size b < n without replacement, recomputes a statistic on each, and uses the sqrt(b)-rescaled subsampling distribution to build a confidence interval and standard error that are stable as b varies.Full
Bayesian Bootstrap (Rubin)Rubin's Bayesian bootstrap for OLS coefficients: draws Dirichlet(1,...,1) weights over the observations and refits weighted least squares each replication to produce a posterior distribution (mean, sd, 95% credible interval) for every coefficient, first-order equivalent to the ordinary bootstrap.Full
BCa Bootstrap IntervalBias-corrected and accelerated (BCa) bootstrap 95% confidence interval for a chosen scalar statistic (mean, median, std, variance, quartiles, IQR) of one numeric variable, correcting the plain percentile interval for median bias (z0) and skewness (jackknife acceleration a).Full
Cluster Bootstrap (OLS)Pairs cluster bootstrap for OLS coefficient inference that resamples whole clusters with replacement and refits to produce cluster-robust bootstrap standard errors and percentile confidence intervals valid under arbitrary within-cluster error correlation.Full
Bootstrap Confidence IntervalFour CI methods (percentile, basic, BCa, studentized) with distribution summary and downsampled samples.Full
Bootstrap RegressionRow-level resampling with OLS per iteration; bootstrap SE, bias, CI, and comparison to classical estimates.Full
Jackknife EstimationLeave-one-out resampling, SE, bias correction, and bias-corrected estimate.Full
Permutation Test (Two-Sample)Exact resampling test for mean/median/t-stat difference with null distribution and two-tailed p-value.Full
Permutation Test (Correlation)Exact test for Pearson/Spearman/Kendall correlation with null distribution and significance.Full
Cross-ValidationK-fold CV (sklearn-compatible) for linear/logistic/ridge/lasso with multiple scoring metrics.Full
Bayesian BootstrapDirichlet-weighted resampling for posterior distribution of a statistic; Bayesian alternative to classical bootstrap.Full

Machine Learning & Choice Models 31

Statios provides machine learning and choice modeling tools: supervised learning with explainability (Random Forest, Gradient Boosting, SVM, Decision Trees, neural networks); model evaluation (cross-validation, ROC/AUC, confusion matrix, feature importance, SHAP); and conjoint analysis (traditional, choice-based, hierarchical Bayes) for preference modeling.

ProcedureWhat it doesStatus
Random ForestFit random forest for classification or regression with feature importances and OOB score.Full
Gradient BoostingFit gradient boosted model (XGBoost if available) with learning curves and feature importances.Full
Support Vector MachineFit SVM for classification or regression with standardized features and support vector counts.Full
Decision TreeFit decision tree with tree rules, feature importances, and depth/leaf metrics.Full
Cross-ValidationK-fold cross-validation with train/test scores and timing metrics across folds.Full
ROC/AUC AnalysisBinary classification ROC curve with AUC and Youden's J optimal threshold.Full
Confusion MatrixMulti-class confusion matrix with per-class precision/recall/F1, normalizations, and macro/weighted averages.Full
Classification MetricsAccuracy, precision, recall, F1, Matthews correlation, Cohen's kappa, and log-loss.Full
Feature ImportanceBuilt-in and permutation-based feature importances with standard errors.Full
SHAP ValuesSHAP values via TreeExplainer (exact TreeSHAP) or KernelExplainer (model-agnostic) with expected values.Limited
SHAP Summary PlotSHAP summary (beeswarm) plot data with feature importance ranking and per-sample contributions.Full
Train/Test PredictFit model on train set and return test predictions, actuals, residuals/probabilities.Full
MLP ClassifierScikit-learn multilayer perceptron for classification with configurable hidden layers and activation.Full
MLP RegressorScikit-learn multilayer perceptron for regression with configurable hidden layers and activation.Full
Deep Neural NetworkKeras/TensorFlow deep network for classification or regression with custom architecture, epochs, batch size.Full
AutoencoderUnsupervised autoencoder for dimensionality reduction and feature learning.Full
CNN for Tabular DataConvolutional neural network adapted for tabular classification or regression.Full
Neural Network Hyperparameter TuningGrid search over neural network hyperparameters with cross-validation.Full
CHAIDChi-squared automatic interaction detection for tree-based classification.Limited
QUESTQuick, unbiased, efficient statistical tree method for classification.Limited
C5.0C5.0 decision tree and rule learner for classification.Limited
Ensemble Trees ComparisonSide-by-side comparison of multiple tree-based ensemble methods.Limited
Tree ExportExport trained tree model in formats suitable for deployment.Limited
Compare TreesComparative metrics across tree models (accuracy, depth, interpretability).Limited
Traditional Conjoint AnalysisRating-based conjoint via OLS with dummy-coded attributes and part-worths.Full
Choice-Based ConjointAggregate multinomial logit conjoint from choice data with part-worths and hit rate.Full
Hierarchical Bayes CBCEmpirical-Bayes iterated-MAP CBC with individual part-worths (not true HB-MCMC).Limited
Market SimulatorMarket share simulation given part-worths and scenarios (logit, first-choice, randomized).Full
Attribute ImportanceRelative attribute importance from part-worth ranges (% allocation).Full
Conjoint DesignFractional experimental design generation (full, orthogonal, D-optimal).Full
Conjoint SegmentationSegment respondents using K-means on individual part-worths.Full

Complex Survey & Missing Data 18

Statios provides complex survey design analysis and missing data handling. Covers design-based inference (Taylor linearization, replicate weights, Rao-Scott corrections) and advanced imputation (MICE, EM, PMM) with proper uncertainty propagation.

ProcedureWhat it doesStatus
Survey design definitionSpecify strata, clusters, sampling weights, and finite population correction; summarize design.Full
Survey-adjusted meansDesign-based estimates with Taylor linearization variance; includes DEFF per variable.Full
Survey-adjusted totalsWeighted population totals with design-adjusted standard errors and DEFFs.Full
Survey proportionsDesign-based category proportions with confidence intervals and design effects.Full
Survey-weighted regressionLinear or logistic regression via WLS with design-adjusted sandwich covariance.Full
Design Effect (DEFF)Ratio of design variance to SRS variance for a mean; includes interpretation guide.Full
Replicate weights analysisJackknife, BRR, or bootstrap SE/CI from replicate weight methodology.Full
Survey t-testTwo-sample design-adjusted t-test on difference of weighted means.Full
Rao-Scott chi-squareChi-square test for categorical association; first-order design effect correction.Full
Missing value summaryComprehensive report of missing counts, patterns, and complete-case counts per variable.Full
Little's MCAR testTest whether missingness is completely random; EM-based ML with chi-square inference.Full
Simple imputationMean, median, or mode imputation for a single variable.Full
KNN imputationImpute missing via k nearest neighbors; standardizes features to avoid scale-dependence.Full
MICE imputationMultiple imputation by chained equations; pooled estimates via Rubin's rules.Full
EM imputationEM algorithm for missing data assuming multivariate normality; returns imputed dataset.Full
Predictive Mean Matching (PMM)Imputation via OLS prediction + nearest-neighbor donor draw; preserves observed distribution.Full
Missing data patternsIdentify missing patterns; check monotone vs non-monotone structure; Little's MCAR suggestion.Full
MNAR sensitivity analysisShift imputed values by delta range; track regression coefficient changes to find tipping point.Full

Data Management, Transforms & Workflow 68

Data management, transforms and workflow procedures in Statios: dataset loading and multi-dataset workflows, direct cell editing with undo/redo, spreadsheet-style formulas for derived variables, variable metadata configuration (types, labels, measurement levels), data transformations (compute, recode, categorize, standardize, filter, lag/lead, dummy), data manipulation (merge, reshape, aggregate, collapse, append, sort, subset, deduplicate, rank, transpose, split/concat columns, pivot), data-to-table formatting, and session/macro recording for reproducible workflows.

ProcedureWhat it doesStatus
Load DatasetLoad data from file (CSV, TSV, XLSX, SPSS .sav, Stata .dta, JSON) into active workspace.Full
Get Dataset InfoRetrieve metadata for active dataset: row/column counts, dtypes, missing cells, columnar status.Full
Get Data WindowFetch windowed rows (e.g., 100 rows at offset 0) for grid rendering.Full
Get Full ColumnsRetrieve whole columns for chart building; large datasets sampled with fixed seed for row alignment.Full
Get VariablesInferred variable metadata (type, measure, labels, value labels) merged with engine-owned overrides.Full
List DatasetsList all open datasets and current active dataset ID.Full
Switch Active DatasetSet a dataset as active (default target for all analyses and edits).Full
Close DatasetRemove a dataset from memory; clears associated undo/redo stacks.Full
Duplicate DatasetCreate independent copy of dataset with all metadata and current state.Full
Rename DatasetChange display name (in-memory; original file unchanged).Full
Export DatasetWrite active dataset to file (CSV, XLSX, SAV, DTA, JSON) with variable metadata.Full
Set Cell ValueEdit single cell; coerces value to column dtype; generates undo entry.Full
Undo Cell EditReverse last cell edit (stack cap 200 per dataset).Full
Redo Cell EditRestore undone cell edit.Full
Insert RowInsert blank row at index; structural edit clears undo stacks.Full
Delete RowRemove row at index; structural edit clears undo stacks.Full
Add VariableAppend new column (all-NaN or default name auto-generated).Full
Delete VariableRemove column; structural edit clears undo stacks.Full
Sort DatasetSort by column(s) ascending/descending; structural edit clears undo stacks.Full
Evaluate FormulaOne-off scalar formula (e.g., =A1+B1); returns value or Excel-style error code.Full
Apply Formula to ColumnFill formula down entire column as derived variable; stored for recomputation.Full
Recompute Derived ColumnsRecalculate all derived columns after data edits in dependency order; reports #REF! for missing inputs.Full
Get FormulasRetrieve stored formula metadata: source, refs (letter→column), anchor row, dependencies.Full
Set Variable TypeConvert column dtype (numeric/string/date/boolean) with honest failure accounting.Full
Set Measurement LevelAssign measure (nominal/ordinal/scale) for SPSS Variable View parity.Full
Set Value LabelsMap raw values to display labels (e.g., 1→'Yes', 0→'No'); empty list clears labels.Full
Auto-Encode CategoricalSPSS-style: text categories → numeric codes with inverse map as value labels.Full
Compute VariableCreate new variable from AST-safe expression (log, sqrt, abs, sin, cos, mean, std, +−*/). Stores result in new column.Full
Recode VariableMap values by pattern (e.g., '1-30'→'low', '31+'→'high') into new column.Full
Standardize VariableTransform to z-scores (mean 0, SD 1), min-max [0,1], or robust (median, IQR).Full
Categorize / BinConvert numeric to categories: equal-width, quantile-based, or custom breaks.Full
Filter Cases (Read-Only)Report which rows match a condition (e.g., age > 30); read-only—use data.subset to drop.Full
Create Dummy VariablesExpand categorical to 0/1 indicators; optionally drop first to avoid collinearity.Full
Lag / LeadShift values k periods forward/backward for time-series predictors.Full
Merge DatasetsSQL-like join (inner/left/right/outer/cross) on key column(s); reports match/unmatch counts.Full
Reshape LongConvert wide → long: multiple value columns become one 'value' + one 'variable' column.Full
Reshape WideConvert long → wide: spread key column values across columns.Full
Aggregate DataGroup by column(s) and compute summaries (mean/sum/count/etc.); produces new dataset.Full
Collapse DataSimilar to aggregate; summarize grouped data (default stats: mean, SD, N).Full
Append DatasetsVertically stack 2+ datasets (row-bind); common columns aligned.Full
Sort DatasetSort by column(s) ascending/descending; optionally replace active or create new dataset.Full
Subset / FilterKeep rows matching condition (AST-safe); replace active dataset.Full
DeduplicateRemove duplicate rows (keep first/last/none); optionally subset columns first.Full
Transpose DatasetFlip rows ↔ columns; optionally use one column as new row index.Full
Split ColumnSplit text column by separator into multiple new columns.Full
Concatenate ColumnsJoin 2+ columns into one (e.g., first + last → fullname).Full
Rank VariableConvert to ranks (1,2,3...); methods: average (ties split), min/max, dense, ordinal.Full
Crosstab CountsBuild contingency table (row × col) from two categorical variables; include marginals.Full
Pivot Longer (Complex)Advanced wide→long: regex on column names extracts multiple groups into separate columns.Full
Save SessionPersist all datasets, variable metadata, and output history to .stios JSON.Full
Load SessionRestore datasets, metadata, and outputs from .stios; repopulates DatasetManager.Full
Session InfoRead session metadata (creation date, version, dataset names, row/col counts).Full
Session DiffCompare two .stios files; report differences in datasets and outputs.Full
Start Macro RecordingBegin recording user actions (transforms, analyses) for later replay.Full
Stop Macro RecordingEnd current macro recording session.Full
Save MacroWrite recorded actions to .sfmacro JSON file.Full
Load MacroRead .sfmacro file; return actions ready to replay.Full
Replay MacroExecute recorded actions in sequence (transform.compute, data.*, etc.) on current dataset.Full
Session UndoReverse a logged action by ID (currently stub; per-dataset cell undo used instead).Limited
Session RedoRestore undone action by ID (currently stub; per-dataset cell redo used instead).Limited
Export ScriptConvert recorded macro actions to Python or R script (reproducible workflow).Full
Regression TableFormat regression results (multiple model specs) in APA/HTML/LaTeX table.Full
Descriptives TableFormat descriptive statistics by groups in APA/HTML/LaTeX table.Full
Correlation TableFormat correlation matrix in APA/HTML/LaTeX; supports Pearson, Spearman, Kendall.Full
Crosstab TableFormat contingency table (observed, expected, residuals) in APA/HTML/LaTeX.Full
Model Comparison TableSide-by-side regression models with fit statistics and significance stars.Full
ANOVA TableFormat one-way/repeated-measures ANOVA table in APA/HTML/LaTeX.Full
Factor Loading TableFormat EFA/PCA loadings with rotation (varimax, promax, etc.).Full

AI, Automation, Graphics & Reporting 19

AI-powered statistical guidance, automated analysis pipelines, publication-quality graphics and report generation, plus a visual rule-based workflow recommender (Guided Path) and a node-based pipeline editor (RLab).

ProcedureWhat it doesStatus
Statios AI assistantAsk the built-in assistant in plain language; it chooses an analysis, runs it on the local engine, and interprets the result. Routed through the Statios relay — your raw data rows never leave your device.Full
Summarize iterationSummarizes executed analyses into key findings as structured JSON (used by the Research Path); routed through the Statios relay.Full
Auto data quality checkSingle-call data quality report: missing patterns, duplicates, type consistency, cardinality anomalies.Full
Auto assumption validationDetect violations of key statistical assumptions (normality, homogeneity, linearity, etc.) for a given test type.Full
Auto exploratory data analysisOne-command EDA: univariate summaries, correlation matrix, key visualizations, and insight extraction.Full
Compare candidate modelsFit multiple regression models (OLS, ridge, lasso, elastic net, robust) on same outcome; rank by IC/CV.Full
HistogramDistribution visualization for a numeric variable with Sturges' bin count and descriptive stats.Full
Scatter plotBivariate X–Y plot with optional grouping and Pearson correlation in response.Full
Box plotMedian, quartiles, and outliers per group (or ungrouped); includes min/max and sample size.Full
Bar chartCategorical value counts or aggregated numeric summaries, with optional value labels.Full
Export statistical tableWrite a result table to DOCX (APA style), LaTeX, HTML, Markdown, or CSV.Full
Generate full reportBuild a report document from analysis sections with optional citations; export as Markdown, HTML, or LaTeX.Full
Build report documentAssemble a deterministic report model from journal entries, dataset metadata, and optional guided plan; handles APA 7 formatting and citation aggregation.Full
Export report to fileRender and write a report model as HTML, Markdown, or BibTeX; optionally saves .bib sidecar.Full
Recommend analysis pathProfile-based deterministic method selector: returns step plan with matched rule, alternatives, and aggregated bibliography.Full
Generate executable plan DAGConvert study profile + matched rule into a directed graph with nodes (analyses), edges (unconditional backbone + conditional branches), and bound parameters.Full
Profile datasetIntrospect active dataset and return partial StudyProfile (data-determined fields only); leaves interview unknowns and variable candidates for UI.Full
AI-guided interview turnConstrained AI conversation that fills StudyProfile fields only via strict-JSON validation; forbids methodological advice.Full
Generate methods sectionRender a plan DAG as AI-authored academic methods prose (intro, phase paragraphs with inline APA 7 citations, alphabetical reference list, BibTeX export); each paragraph is citation-validated against the MKB registry with a template-based fallback. Replaces the earlier template-only justification.Full