← ClaudeAtlas

quant-multi-group-deplisted

Compare an outcome across three or more dependent (within-subject) measurements — e.g., three timepoints, three conditions rated by the same respondents, or three brand dimensions evaluated within-person. Uses a permutation Friedman test as the omnibus, followed by MANDATORY pairwise post-hoc sign-flip permutation tests with BH FDR correction and Cohen's dz for each pair. Triggers: "three timepoints", "three conditions", "repeated measures", "within-subject", "same respondents across all", "before during after", "track over time", "multiple waves", "repeated measurements", "friedman test", "3+ paired conditions", "longitudinal comparison".
JSerek/quant-skills · ★ 0 · AI & Automation · score 70
Install: claude install-skill JSerek/quant-skills
# quant-multi-group-dep — Three or More Dependent Measurements ## Objective Test whether an outcome variable differs across 3 or more within-subject conditions or timepoints, and identify which specific conditions differ and by how much. Each respondent must provide a measurement for every condition (complete repeated design). --- ## Pre-flight **Required inputs:** - `output/data/cleaned_data.csv` - `output/data/column_metadata.json` **Requirements:** - 3–10 numeric/ordinal columns representing the repeated conditions - OR: one outcome column + one condition variable (3+ levels) + respondent ID (long format) - Minimum n: 20 complete cases (all conditions observed per respondent) - If exactly 2 conditions: redirect to `/quant-two-group-dep` --- ## Pipeline ### Step 1 — Load data and metadata ```python # See references/pipeline.py → load_prepared_data() ``` --- ### Step 2 — AskUserQuestion: data structure ``` AskUserQuestion: question: "How are your repeated measurements structured?" header: "Data structure" options: - label: "Wide format — one column per condition/timepoint (Recommended)" description: "E.g., 'satisfaction_t1', 'satisfaction_t2', 'satisfaction_t3' as separate columns." - label: "Long format — one outcome column + condition column" description: "E.g., 'satisfaction' column with a 'timepoint' column. Requires a respondent ID column for pairing." ``` --- ### Step 3 — AskUserQuestion: select columns **Wid