Pg analysis approval

Skill Exekiel179/pingouin-psych-stats/skills/pg-analysis-approval

Review and approve Pingouin psychology analyses before final reporting. Use after data screening or statistical output and before interpretation to catch common problems: wrong test, ignored repeated measures, missing assumption checks, multiple-comparison errors, unsupported causal claims, incomplete reporting, or output-format mismatches.From its SKILL.md

Install

npx -y skills add Exekiel179/pingouin-psych-stats --skill pg-analysis-approval

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

0 stars0 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

SKILL.md

2.7 KB, 549 tokens by cl100k_base, as published. Nobody here has run it

PG Analysis Approval

Use as the final statistical checkpoint before writing conclusions or deliverables.

Load

Read:

../../references/supervision-gates.md
../../references/workflow-index.md
../../references/apa-output-template.md if checking report prose.
../../references/archive-contract.md when reviewing an archived run.

Read ../../references/pingouin-api-quickref.md only if an API/function choice is questionable.

Inputs

Review whatever is available:

User request and research question.
Dataset screening summary.
Pingouin code.
Pingouin result tables.
Draft interpretation/report.
Requested output format and figure/table requirements.

Approval Checklist

Check:

S0 Scope: design, outcome scale, unit of analysis, dependency structure.
S1 Data: columns, missingness, group sizes, repeated IDs, long/wide shape.
S2 API: current Pingouin function signature, no deprecated function, correct parameters.
S3 Assumptions: normality/homogeneity/sphericity/linearity/independence as relevant.
S4 Result: values come from output; df, p, CI, effect size, correction are present when needed.
S5 Interpretation: no unsupported causal, clinical, or construct-level overclaim.
Deliverable: requested format, language, tables, figures, and reproducibility needs are addressed.
Archive: exact code, numerical output, report source, and audit record are stored under one run directory.

Decision Labels

Use one label:

APPROVED: ready to report.
APPROVED_WITH_NOTES: acceptable but caveats must be included.
REVISE: fixable issue before reporting.
BLOCKED: missing data/design information or wrong model family.

Output Format

Decision: <APPROVED / APPROVED_WITH_NOTES / REVISE / BLOCKED>
Critical issues: <none or concise list>
Required fixes: <none or exact actions>
Reporting notes: <effect size, CI, correction, caveat, format>
Audit: S0 <status>; S1 <status>; S2 <status>; S3 <status>; S4 <status>; S5 <status>.

Hard Stops

Return BLOCKED if:

Outcome scale or dependency structure makes the selected test invalid.
Repeated/nested data were analyzed as independent rows.
A causal mediation/causal conclusion is requested from unsuitable data without caveat.
Required numerical output is absent but the user asks for final result prose.
The correct method is outside Pingouin's scope.

What ships with it: 1 file

264 B alongside SKILL.md

agents/

openai.yaml264 B