How do I install the Pg Multivariate skill?

Run npx skillmds@latest add exekiel179/pg-multivariate in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Pg Multivariate safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Pg Multivariate?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Pg Multivariate free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Pg Multivariate?

Exekiel179 (@exekiel179) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Pg Multivariate by exekiel179 — Agent Skill

Name: Pg Multivariate
Author: Exekiel179

Pg Multivariate

Run or generate Pingouin code for multivariate comparisons — Hotelling's T-squared test (multivariate_ttest) — plus the multivariate assumption checks box_m (equal covariance) and multivariate_normality, for designs with several dependent variables.

Exekiel179 efc72f4 2 files · 3.0 KB Updated Sep 17, 2026

File contents

PG Multivariate

Use when several continuous dependent variables are compared together (a mean-vector / profile question), rather than one outcome at a time.

Load

Read:

../../references/supervision-gates.md
../../references/pingouin-api-quickref.md
../../references/apa-output-template.md if writing results.

Decision Rules

Compare one group's mean vector to a reference (one-sample) -> pg.multivariate_ttest(X).
Compare mean vectors of two independent groups -> pg.multivariate_ttest(X, Y).
Paired multivariate comparison (same participants) -> pg.multivariate_ttest(X, Y, paired=True).
Check equal covariance matrices across groups (assumption) -> pg.box_m(data, dvs, group).
Check multivariate normality (assumption) -> pg.multivariate_normality(X).

Required Inputs

Two or more continuous dependent variables (the outcome vector).
Grouping column for the two-sample test and Box's M.
Whether the comparison is one-sample, independent, or paired.
X/Y are arrays or frames of shape (n, k); drop rows with missing DVs first.

Code Patterns

Hotelling's T-squared, two independent groups:

X = df.loc[df["group"].eq("A"), ["v1", "v2", "v3"]].to_numpy()
Y = df.loc[df["group"].eq("B"), ["v1", "v2", "v3"]].to_numpy()
res = pg.multivariate_ttest(X, Y).round(3)
pg.print_table(res)

One-sample (mean vector vs zero; subtract a reference first if needed):

res = pg.multivariate_ttest(df[["v1", "v2", "v3"]].to_numpy()).round(3)
pg.print_table(res)

Assumption checks:

pg.box_m(df, dvs=["v1", "v2", "v3"], group="group")           # equal covariance
pg.multivariate_normality(df[["v1", "v2", "v3"]], alpha=0.05)  # Henze-Zirkler

Output Checks

multivariate_ttest returns index hotelling with T2, F, df1, df2, pval. box_m returns Chi2, df, pval, equal_cov (bool). multivariate_normality returns (hz, pval, normal); normal=False warns against assuming multivariate normality.

Guardrails

Hotelling's T-squared needs n greater than the number of DVs and roughly multivariate-normal data; check multivariate_normality.
A significant Box's M (unequal covariance) undermines the pooled test; report it.
Pingouin covers Hotelling's T-squared, not full factorial MANOVA — for that recommend statsmodels or R.
Report T2, F, df, p, and which assumption checks were run.
End result-bearing answers with one compact S0-S5 audit line.

---name: pg-multivariatedescription: Run or generate Pingouin code for multivariate comparisons — Hotelling's T-squared test (multivariate_ttest) — plus the multivariate assumption checks box_m (equal covariance) and multivariate_normality, for designs with several dependent variables.---# PG MultivariateUse when several continuous dependent variables are compared together (a mean-vector / profile question), rather than one outcome at a time.## LoadRead:- `../../references/supervision-gates.md`- `../../references/pingouin-api-quickref.md`- `../../references/apa-output-template.md` if writing results.## Decision Rules- Compare one group's mean vector to a reference (one-sample) -> `pg.multivariate_ttest(X)`.- Compare mean vectors of two independent groups -> `pg.multivariate_ttest(X, Y)`.- Paired multivariate comparison (same participants) -> `pg.multivariate_ttest(X, Y, paired=True)`.- Check equal covariance matrices across groups (assumption) -> `pg.box_m(data, dvs, group)`.- Check multivariate normality (assumption) -> `pg.multivariate_normality(X)`.## Required Inputs- Two or more continuous dependent variables (the outcome vector).- Grouping column for the two-sample test and Box's M.- Whether the comparison is one-sample, independent, or paired.- `X`/`Y` are arrays or frames of shape (n, k); drop rows with missing DVs first.## Code PatternsHotelling's T-squared, two independent groups:```pythonX = df.loc[df["group"].eq("A"), ["v1", "v2", "v3"]].to_numpy()Y = df.loc[df["group"].eq("B"), ["v1", "v2", "v3"]].to_numpy()res = pg.multivariate_ttest(X, Y).round(3)pg.print_table(res)```One-sample (mean vector vs zero; subtract a reference first if needed):```pythonres = pg.multivariate_ttest(df[["v1", "v2", "v3"]].to_numpy()).round(3)pg.print_table(res)```Assumption checks:```pythonpg.box_m(df, dvs=["v1", "v2", "v3"], group="group") # equal covariancepg.multivariate_normality(df[["v1", "v2", "v3"]], alpha=0.05) # Henze-Zirkler```## Output Checks`multivariate_ttest` returns index `hotelling` with `T2`, `F`, `df1`, `df2`, `pval`. `box_m` returns `Chi2`, `df`, `pval`, `equal_cov` (bool). `multivariate_normality` returns `(hz, pval, normal)`; `normal=False` warns against assuming multivariate normality.## Guardrails- Hotelling's T-squared needs n greater than the number of DVs and roughly multivariate-normal data; check `multivariate_normality`.- A significant Box's M (unequal covariance) undermines the pooled test; report it.- Pingouin covers Hotelling's T-squared, not full factorial MANOVA — for that recommend statsmodels or R.- Report T2, F, df, p, and which assumption checks were run.- End result-bearing answers with one compact S0-S5 audit line.

Frequently asked questions