quant-correlationlisted

# quant-correlation ## Objective Quantify pairwise associations between two or more variables, with: - Permutation-based significance (no normality assumption) - Bootstrap 95% CIs on each correlation coefficient - Effect size interpretation (correlation as effect size: negligible/small/medium/large) - Interactive heatmap + scatter matrix - Plain-language interpretation of the pattern **When to use this skill:** - Exploring which variables move together before modelling - Checking for multicollinearity between predictors - Providing a correlation matrix in a research report - Post-analysis: understanding what drives a key outcome variable **When NOT to use:** - You want to predict one variable from others → use `quant-model-continuous` or `quant-model-ordinal` - You have only 2 variables and want a significance test → simpler to use `quant-two-group-ind` or note correlation is equivalent - You have nominal (unordered categorical) variables → use Cramér's V instead (not currently in scope) --- ## Pre-flight checks 1. **Minimum variables:** ≥ 2 numeric or ordinal columns selected 2. **Minimum n:** ≥ 10 pairwise complete observations for any pair; warn at n < 30 3. **Variable type check:** Skip purely nominal columns and warn user 4. **Scale check:** Ordinal variables (e.g., Likert 1–5) → recommend Spearman or Kendall's tau-b 5. **Missing data:** Pairwise complete case analysis by default; note pairwise n per cell if it varies --- ## AskUserQuestion protocol ### Step 1