quant-model-ordinallisted

# quant-model-ordinal ## 1. Objective Fit a **Cumulative Link Model (CLM)** — also known as proportional odds logistic regression — for an **ordinal dependent variable** (typically a Likert 1–5 or 1–7 scale). The model estimates the log-odds of being at or above each category threshold, assuming the effect of each predictor is constant across all thresholds (proportional odds assumption). This is tested formally and flagged if violated. **When to use:** - DV is ordinal: Likert 1–5, satisfaction ratings, NPS bands, agreement scales - Goal: understand which predictors shift respondents toward higher (or lower) categories - N per category ≥ 5 (sparse categories flagged with warning) **When NOT to use:** - DV is continuous or near-continuous (many unique values) → `quant-model-continuous` - DV is binary (0/1) → `quant-model-nominal` - DV has 3+ unordered categories → `quant-model-multinomial` - Data is nested/hierarchical → `quant-model-multilevel` --- ## 2. Pre-flight checks | Check | Threshold | Action | |---|---|---| | DV categories | ≥ 3 ordered categories | Error if < 3 | | Min n per category | ≥ 5 | Warn user, proceed with caution | | Complete cases | ≥ 30 after listwise deletion | Error if < 10 | | Proportional odds | Score test p > .05 | Flag 🟡/🔴, suggest partial PO model | --- ## 3. AskUserQuestion protocol ### Q1 — Input file ``` What is the path to your input CSV file? ``` ### Q2 — Outcome variable ``` Which column is the outcome (dependent) variable? Thi