stat-research-pipeline

End-to-end applied statistical research pipeline. Takes a research question and dataset, runs literature review, statistical analysis with parallel method tracks, and produces a complete manuscript (Markdown + LaTeX/PDF). Use when user says "application pipeline", "applied analysis", "analyze my data and write paper", "end-to-end analysis", or wants to go from raw data to manuscript. Covers all outcome types: continuous, binary, ordinal, nominal, count, survival, repeated measures, time series, multivariate, DOE, meta-analysis, and SEM. Human-in-the-loop by design: the skill handles the standardized workflow, while the user owns study framing, high-stakes judgment, and final sign-off.

Server-side extension that completes the full analysis pipeline for binary outcome variables after vera-data-binary-reviewing has run. Adds remaining association tests (chi-square/Fisher's for additional predictors, point- biserial for continuous predictors), stratified odds ratio analysis with Breslow-Day test and forest plot, full modeling (logistic regression with OR and 95% CI, Hosmer-Lemeshow GOF, pseudo-R2, ROC curve with AUC, classification table, tree-based classification with CART/RF/GBM), and cross-method variable importance comparison on a 0-100 unified scale. Generates manuscript-ready methods.md and results.md with formatted tables, publication-quality figures, and references.bib. Applies output variation and code style variation for natural, non-repetitive output. Triggered after vera-data-binary-reviewing completes and its PART 0–2 artifacts (code.R, code.

Runs class balance diagnostics and primary association tests for binary outcome variables (0/1, yes/no, survived/died, pass/fail). Produces proportion tables, class balance check with rare-event warning, descriptives by outcome level, chi-square test of independence with Cramer's V, Fisher's exact test when cell counts are small, odds ratio with 95% CI, and a mosaic or grouped bar chart. Outputs .R and .py scripts with 2 publication-quality plots. Triggered when user has a binary/dichotomous outcome and says "binary outcome," "survived or died," "yes or no," "pass or fail," "0/1," "classification," "binary DV," or names a binary variable like survived, admitted, defaulted, churned, diagnosed. Does not handle continuous, count, survival time, ordinal, repeated measures, or SEM.

Full CFA skill for measurement-model evaluation. Extends the initial CFA fit with reliability, convergent/discriminant validity, alternative-model comparison, modification-index review, subgroup invariance testing, and manuscript-ready methods/results. Trigger after vera-data-cfa-reviewing completes and its PART 0–2 artifacts are present (see ../../CROSS-SKILL-INTERFACE.md). If invoked directly without those artifacts, halts and prompts the user to run testing first or supply equivalent PART 0–2 code.

Complete skill family for confirmatory factor analysis (CFA) and measurement model evaluation. Use when the user asks for CFA, latent factor validation, measurement models, factor loadings, convergent/discriminant validity, or measurement invariance. Uses a two-step structure: testing for initial diagnostics, analyzing for full pipeline.

Runs the CFA pipeline: collect indicator structure, check basic identification and estimator requirements, fit a primary CFA model, and report fit indices plus standardized loadings with a recommendation block for full SEM analysis. Trigger when the user asks for CFA, confirmatory factor analysis, measurement model testing, latent factor validation, factor loading assessment, construct validity, scale validation, measurement invariance check, or psychometric evaluation. Does not cover full structural path modeling (use vera-data-path-reviewing) or longitudinal change modeling (use vera-data-growth-reviewing).

Server-side extension that completes the full analysis pipeline for continuous outcome variables after vera-data-continuous-reviewing has run. Adds remaining hypothesis tests (ANOVA, post-hoc, nonparametric), subgroup analysis with interaction tests and forest plots, full modeling (OLS with diagnostics, quantile regression, tree-based exploratory), and model comparison. Generates manuscript-ready methods.md and results.md with formatted tables, publication-quality figures, and references.bib. Applies output variation and code style variation for natural, non-repetitive output. Triggered after vera-data-continuous-reviewing completes and its PART 0–2 artifacts are present (see ../../CROSS-SKILL-INTERFACE.md). If invoked directly without those artifacts, halts and prompts the user to run testing first or supply equivalent PART 0–2 code.

Runs distribution diagnostics and primary hypothesis tests for continuous outcome variables. Produces Shapiro-Wilk normality check, skewness, kurtosis, Q-Q plot, and one fully interpreted group comparison (Welch's t for 2 groups or ANOVA with Tukey HSD for 3+ groups) with effect sizes and nonparametric confirmation. Ends with a recommendation block listing Outputs .R and .py scripts with 2 publication-quality plots. Triggered when user has a continuous/numeric outcome and says "analyze continuous outcome," "my DV is numeric," "compare group means," or names a continuous variable like weight, score, income, time, cost, mpg, blood pressure. Does not handle binary, count, survival, ordinal, repeated measures, or SEM outcomes.

Server-side extension that completes the full analysis pipeline for count outcome variables after vera-data-count-reviewing has run. Adds remaining hypothesis tests (additional group comparisons, continuous predictor correlations), subgroup analysis with stratified rate ratios and interaction tests, full modeling (Poisson, Negative Binomial, Zero-Inflated Poisson, Zero-Inflated Negative Binomial, Hurdle, tree-based exploratory), and model comparison with unified variable importance. Generates manuscript-ready methods.md and results.md with formatted tables, publication-quality figures, and references.bib. Applies output variation and code style variation for natural, non-repetitive output. Triggered after vera-data-count-reviewing completes and its PART 0–2 artifacts are present (see ../../CROSS-SKILL-INTERFACE.md). If invoked directly without those artifacts, halts and prompts the user to run testing first or supply equivalent PART 0–2 code.

Runs distribution diagnostics and primary hypothesis tests for count outcome variables (non-negative integers with no upper bound). Produces frequency table, overdispersion assessment, bar chart of count distribution with Poisson overlay, and one fully interpreted group comparison (Poisson/NB rate test or regression with LR test) with incidence rate ratios. Detects zero-inflation and overdispersion, flags for zero-inflated models in full analysis. Outputs .R and .py scripts with 2 publication-quality plots. Triggered when user has a count/frequency outcome and says "count outcome," "number of events," "Poisson," "incidents," "rate," "per capita." Does not handle binary, continuous, survival, ordinal, repeated measures, or SEM outcomes.

Server-side extension that completes the full analysis pipeline for designed experiments (DOE) after vera-data-doe-reviewing has run. Adds simple effects analysis, contrast analysis, effect magnitude ranking, response surface methodology (RSM) with contour plots and canonical analysis, fractional factorial alias structure, split-plot error terms, residual diagnostics, optimal factor settings via desirability function, and tree-based variable importance (RF + LightGBM). Generates manuscript-ready methods.md and results.md with formatted tables, publication-quality figures, and references.bib. Applies output variation, code style variation for natural, non-repetitive output. Triggered after vera-data-doe-reviewing completes and its PART 0–2 artifacts are present (see ../../CROSS-SKILL-INTERFACE.md). If invoked directly without those artifacts, halts and prompts the user to run testing first or supply equivalent PART 0–2 code.

Runs distribution diagnostics and primary analysis for designed experiments (DOE) with controlled factors. Produces cell descriptives, balance check, residual normality, Levene's test, full factorial ANOVA (Type III SS) with partial eta-squared, interaction plots, and Tukey HSD / simple effects post-hoc. Outputs .R and .py scripts with 2 publication-quality plots. Triggered when user mentions "experimental design," "DOE," "factorial design," "treatment effects," "blocking," "randomized experiment," "split-plot," "response surface," "2^k factorial," "main effects and interactions," "Latin square," "fractional factorial," "CRD," "RCBD," or describes manipulated factors. Fractional factorial and RSM are identified and flagged; full design-resolution analysis delegated to vera-data-doe-generating. Does not handle observational studies, repeated measures, or SEM outcomes.

Full longitudinal SEM skill for latent growth and change models. Extends the initial fit with nonlinear trajectories, latent-basis and change-score variants, subgroup comparisons, parallel-process models when available, and manuscript-ready output. Trigger after vera-data-growth-reviewing completes and its PART 0–2 artifacts are present (see ../../CROSS-SKILL-INTERFACE.md). If invoked directly without those artifacts, halts and prompts the user to run testing first or supply equivalent PART 0–2 code.

Complete skill family for longitudinal SEM focused on latent growth and change. Use when the user asks for latent growth models, latent change score models, nonlinear change, repeated latent trajectories, or longitudinal SEM in general. Uses a two-step structure: testing for initial diagnostics, analyzing for full pipeline.

Longitudinal SEM skill for latent growth and change. Collects wave structure, checks whether a growth/change model is identified, fits an initial linear growth or latent change specification, and reports fit plus trajectory parameters with a recommendation block. Trigger when the user asks for latent growth models, growth curves in SEM, latent change score models, longitudinal latent trajectories, growth curve modeling, LGM, LCSM, trajectory analysis, or change over time in SEM. Does not handle cross- sectional CFA (use vera-data-cfa-reviewing) or non-longitudinal structural models (use vera-data-path-reviewing).

Server-side extension that completes the full analysis pipeline for meta-analysis after vera-data-meta-reviewing has run. Adds publication bias assessment (funnel plot, Egger's test, trim-and-fill), sensitivity analysis (leave-one-out, influence diagnostics, cumulative meta-analysis), subgroup analysis with Q_between, meta-regression with bubble plots, advanced modeling (REML, Knapp-Hartung, Bayesian, three-level), and model comparison. Generates manuscript-ready methods.md and results.md with formatted tables, publication-quality figures, and references.bib. Applies output variation and code style variation for natural, non-repetitive output. Triggered after vera-data-meta-reviewing completes and its PART 0–2 artifacts are present (see ../../CROSS-SKILL-INTERFACE.md). If invoked directly without those artifacts, halts and prompts the user to run testing first or supply equivalent PART 0–2 code.

Runs study-level diagnostics and primary pooled estimation for meta-analysis of summary statistics across multiple studies. Produces forest plot, heterogeneity assessment (Q, I-squared, tau-squared), pooled estimates under both fixed-effects and random-effects models, and prediction interval. Ends with a recommendation block listing Outputs .R and .py scripts with 1 publication-quality forest plot. Triggered when user says "meta-analysis," "systematic review," "pooled estimate," "combine studies," "forest plot," "heterogeneity," "I-squared," "effect size synthesis," or "aggregate results across studies." Does not handle network meta-analysis or individual participant data.

End-to-end statistics methodology research pipeline. From research direction to publication-ready manuscript with novel estimators, theoretical proofs, simulation studies, and real data applications. Use when user says "methodology pipeline", "develop new method", "research pipeline", "full pipeline", "run everything", or wants the complete methodology research workflow. Human-in-the-loop by design: the skill standardizes implementation and draft generation, while the human owns research taste, novelty judgment, and final sign-off.

Server-side extension that completes the full analysis pipeline for multiple continuous outcome variables after vera-data-multivariate-reviewing has run. Adds follow-up univariate ANOVAs with Bonferroni-corrected pairwise comparisons, full discriminant function analysis with classification accuracy, MANCOVA and two-way MANOVA, profile analysis (parallelism, equal levels, flatness), canonical correlation analysis, PCA for dimension reduction, multivariate multiple regression, and tree-based importance comparison across DVs. Synthesizes insights across methods and generates manuscript-ready methods.md and results.md with formatted tables, publication-quality figures, and references.bib. Applies output variation and code style variation for natural, non-repetitive output. Triggered after vera-data-multivariate- reviewing completes and its PART 0–2 artifacts are present (see ../../../..

Multivariate diagnostics and primary hypothesis tests for multiple continuous outcomes analyzed simultaneously. Produces multivariate normality (Mardia, Henze-Zirkler), Box's M test, scatterplot matrix, correlation matrix, and one multivariate group comparison (Hotelling's T-squared for 2 groups or one-way MANOVA with all four test statistics for 3+ groups) with partial eta-squared per DV and univariate follow-up ANOVAs. Ends with a recommendation block. Outputs .R and .py scripts with publication-quality plots. Handles group- difference tests on 2+ continuous dependent variables (Hotelling's T², MANOVA, one-way MANOVA). Dimensionality-reduction methods (PCA, factor analysis) and canonical correlation are delegated to vera-data-multivariate-generating.

Server-side extension that completes the full analysis pipeline for nominal (unordered multi-class) outcome variables after vera-data-nominal-reviewing has run. Adds remaining association tests (Chi-square/Fisher's for categorical, ANOVA/Kruskal-Wallis for continuous predictors, pairwise class comparisons), subgroup analysis with interaction tests, full modeling (multinomial logistic regression with RRR, LDA, CART, Random Forest, LightGBM multi-class), confusion matrices, unified variable importance (0-100), and cross-method insight synthesis. Generates manuscript-ready methods.md and results.md with formatted tables, publication-quality figures, and references.bib. Applies output variation and code style variation for natural, non-repetitive output. Triggered after vera-data-nominal-reviewing completes and its PART 0–2 artifacts are present (see ../../../../CROSS-SKILL- INTERFACE.md).

Runs class distribution diagnostics and primary association tests for nominal (unordered multi-class) outcome variables. Produces class frequency table, proportions, rare-class warnings, bar chart, and one interpreted association test (Chi-square/Cramer's V for categorical predictors, or ANOVA/Kruskal-Wallis for continuous predictors) with effect sizes. Outputs .R and .py scripts with 2 publication-quality plots. Triggered when user has a nominal outcome with 3+ unordered levels and says "nominal outcome," "multi-class," "unordered categories," "multinomial," or names a variable like species, diagnosis type, region. Does NOT handle binary (2-level), ordinal, continuous, count, survival, or SEM outcomes.

Server-side extension that completes the full analysis pipeline for ordinal outcome variables after vera-data-ordinal-reviewing has run. Adds nonparametric tests, subgroup analysis, and dual-path modeling: Path A ignores ordering (multinomial logistic, CART, RF, LightGBM) while Path B respects ordering (proportional odds with Brant test, adjacent-category logit, continuation-ratio logit, stereotype model, ordinal-aware trees and LightGBM). Cross-path synthesis compares variable importance rankings across both paths. Generates manuscript-ready methods.md and results.md with formatted tables, publication-quality figures, and references.bib. Applies output variation, code and code style variation for natural, non-repetitive output. Triggered after vera-data-ordinal-reviewing completes and its PART 0–2 artifacts are present (see ../../CROSS-SKILL-INTERFACE.md).

Runs distribution diagnostics and primary hypothesis tests for ordinal outcome variables. Produces frequency tables, cumulative proportions, ordinal bar chart, and one fully interpreted nonparametric group comparison (Mann-Whitney U for 2 groups or Kruskal-Wallis for 3+ groups) with effect sizes (rank- biserial r or Cliff's delta) and Jonckheere-Terpstra trend test. Ends with a recommendation block listing additional analyses available. Outputs .R and .py scripts with 2 publication-quality plots. Triggered when user has an ordinal outcome and says "ordinal outcome," "Likert scale," "ordered categories," "rating scale," "severity levels," "none/mild/moderate/ severe," "low/medium/high," "improvement levels," "ranked outcome," "ordered factor," or names an ordinal variable like satisfaction, severity, agreement, stage, grade, rating.

Full SEM analysis skill. Extends the initial structural model with indirect effects, alternative-path comparison, multigroup testing, residual review, and manuscript-ready methods/results. Trigger after vera-data-path-reviewing completes and its PART 0–2 artifacts are present (see ../../CROSS-SKILL-INTERFACE.md). If invoked directly without those artifacts, halts and prompts the user to run testing first or supply equivalent PART 0–2 code.

Complete skill family for full structural equation modeling that combines measurement and structural paths. Use when the user asks for SEM, latent path models, latent mediation, structural paths among latent variables, or a full theory-driven SEM rather than CFA alone.

Structural equation modeling skill. Collects latent constructs and path structure, checks identification and estimator requirements, fits one primary SEM, and reports global fit plus key structural paths with a recommendation block for advanced analyses. Trigger when the user asks for SEM, latent path analysis, full structural equation modeling, path coefficients, structural paths, mediation model, latent variable model, causal SEM, direct and indirect effects, or path diagram. Does not handle CFA-only models (use vera-data-cfa-reviewing) or longitudinal growth models (use vera-data-growth-reviewing).

Server-side extension that completes the full analysis pipeline for repeated measures / longitudinal designs with a continuous outcome after vera-data- repeated-reviewing has run. Adds pairwise comparisons at each time point and within each group, simple effects analysis, subgroup three-way interactions, linear mixed models (random intercept, random slope, growth curve), GEE with multiple correlation structures, tree-based exploratory analysis on subject- level features, and model comparison with unified variable importance (0-100). Generates manuscript-ready methods.md and results.md with formatted tables, publication-quality figures, and references.bib. Applies output variation and code style variation for natural, non-repetitive output. Triggered after vera- data-repeated-reviewing completes and its PART 0–2 artifacts are present (see ../../CROSS-SKILL-INTERFACE.md).

Trajectory diagnostics and primary hypothesis tests for repeated measures / longitudinal designs with a continuous outcome. Produces spaghetti plots with group mean ribbons, ICC, descriptives per time point, attrition check, and one interpreted test: paired t-test for 2 time points, repeated measures ANOVA for 3+ time points in one group, or mixed ANOVA (time x group) for 3+ time points with 2+ groups, including Mauchly sphericity and Greenhouse-Geisser correction. Ends with a recommendation block. Outputs .R and .py scripts with publication-quality plots. Trigger when user has a continuous outcome measured on the same subjects over time and says "repeated measures," "longitudinal," "within-subjects," "pre-post," "paired," "panel data," "growth curve," "multilevel," "mixed model," or "correlated observations." Does not handle binary, count, survival, ordinal, or cross-sectional outcomes.

Server-side extension completing the full analysis pipeline for right-censored survival outcomes after vera-data-survival-reviewing has run. Adds univariate Cox screening, subgroup analysis with stratified Cox and interaction tests, full Cox Proportional Hazards with Schoenfeld diagnostics, time-varying covariate Cox, recurring event models (Andersen-Gill, PWP, frailty), Accelerated Failure Time models (Weibull, log-normal, log-logistic), Random Survival Forest, and gradient boosting survival. Produces unified variable importance (0-100 scale), manuscript-ready methods.md and results.md with formatted tables, publication figures, and references.bib. Applies output variation and code style variation for natural, non-repetitive output. Triggered after vera-data-survival-reviewing completes and its PART 0–2 artifacts are present (see ../../CROSS-SKILL-INTERFACE.md).

Runs survival diagnostics and primary hypothesis tests for right-censored time-to-event outcomes. Produces Kaplan-Meier curves with number-at-risk tables, log-rank test for group comparison, median survival, landmark survival rates, and HR preview from univariate Cox regression. Outputs .R and .py scripts with publication-quality plots. Triggered when user has a time-to-event outcome and says "survival outcome," "time to event," "Kaplan-Meier," "hazard," "OS/PFS/DFS." Right-censored data only. Competing risks, recurrent events, and multi-state models are out of scope. Does not handle binary, count, continuous, ordinal, repeated measures, or SEM outcomes.

Server-side extension that completes the full analysis pipeline for time series data after vera-data-timeseries-reviewing has run. Adds SARIMA, exponential smoothing (ETS/Holt-Winters), structural break tests, Granger causality, subseries and rolling window analysis, regime detection, full classical model suite (ARIMA, SARIMA, ETS, GARCH, VAR, spectral, regression with ARIMA errors), ML-based forecasting (RF + LightGBM on lagged features), and cross-method forecast comparison. Generates manuscript-ready methods.md and results.md with formatted tables, publication-quality figures, and references.bib. Applies output variation and code style variation for natural, non-repetitive output. Triggered after vera-data-timeseries-reviewing completes and its PART 0–2 artifacts are present (see ../../CROSS-SKILL-INTERFACE.md). If invoked directly without those artifacts, halts and prompts the user to run testing first or supply equivalent PART 0–2 code.

Runs time series diagnostics and primary ARIMA modeling for temporal data. Produces time plot, ACF/PACF, seasonal decomposition, stationarity tests (ADF + KPSS), and one fully interpreted ARIMA model with forecast and prediction intervals. Identifies trend, seasonality, and non-stationarity; fits ARIMA/SARIMA as baseline. Advanced methods (Prophet, state-space, VAR) delegated to vera-data-timeseries-generating. Outputs .R and .py scripts with 2 publication-quality plots. Triggered when user has temporal/time series data and says "time series," "temporal data," "forecast," "ARIMA," "seasonal," "trend," "autocorrelation," "monthly data," "daily data," "quarterly," "stationarity," or names a time-indexed variable like sales over time, temperature, stock price, monthly passengers. Does not handle panel/longitudinal data (redirect to vera-data-repeated-reviewing, or vera-data-repeated-generating for growth-curve / mixed-model extensions), cross-sectional data, or spatial time series.