Skill guide: /stats — design, assumptions, method, diagnostics, effects

statistics is the hypothesis-testing skill. It never starts with a test: it identifies the design, audits the assumptions, then chooses the method, runs diagnostics and reports effects with uncertainty.

How to call it

text

/stats Using pss_trial.xlsx, test whether the biofeedback group shows a larger PSS-10 reduction than control. Check assumptions, choose the appropriate test, and report the effect size with a 95% CI and achieved power.

What happens

Design first: groups, repeated measures, covariates, the scale of each variable.
Assumptions: normality, variance homogeneity, sphericity, independence — reported with the numbers.
Method: the test or model, justified; parametric ↔ non-parametric switches are explained.
Diagnostics and effects: effect sizes with CIs, achieved power and design sensitivity, multiple-comparison control.

The statistics turn: dataset loaded, code run in the sandbox, report on the right.

The summary: test statistics, effect sizes with CIs, achieved power and design sensitivity.

Method skills it composes

ttest · anova · power · bayesian (Bayes factor) · stats_tests (parametric ↔ non-parametric selection) · assumptions · effect_size · multiple_comparisons
reliability (α · ω) · efa · cfa · sem · irt

Reliability with bootstrap intervals, then EFA and CFA — the psychometric skills in one pass.

The CFA measurement model drawn with loadings and fit indices.

What you get

The statistical report (Markdown) and a structured result (JSON) the writing step quotes from.
Plots (power, group comparisons, path diagrams) as editable artifacts.

Tips

Ask for what a reviewer will ask for — "effect size with a 95% CI and achieved power" — and it is in the report.
On Deep effort an independent review agent re-checks assumptions, multiple comparisons and sample size before the result is shown.