quant-model-nominallisted

# quant-model-nominal — Binary Logistic Regression ## 1. Objective Fit a binary logistic regression model predicting a binary (0/1) outcome from one or more predictors. Uses HC3 heteroscedasticity-consistent robust standard errors via the `sandwich` package. Outputs odds ratios, fit indices (Nagelkerke R², McFadden R², AUC), ROC curve data, confusion matrix, assumption flags, and a dual-tab HTML report. **Route here when:** - The dependent variable has exactly 2 unique values (0/1, yes/no, aware/not, converted/not) - Predictors are continuous, binary, or categorical (factor) --- ## 2. Pre-flight Required columns in the input CSV: - `outcome_col`: binary (must have exactly 2 unique non-missing values) - `predictor_cols`: one or more columns (numeric or categorical) Minimum n: 30 complete cases (20 per predictor recommended; more is always better). --- ## 3. AskUserQuestion protocol ### Q1 — Input file ``` Which file contains your data? Options: [user provides path] ``` ### Q2 — Outcome column ``` Which column is your binary outcome? (It must have exactly 2 unique values, e.g., 0/1, "Yes"/"No", "Aware"/"Not aware".) ``` ### Q3 — Predictor columns ``` Which columns are your predictors? Select all that apply. [list columns from the dataset] ``` ### Q4 — Separation (conditional on separation flag 🔴) ``` Complete or near-complete separation detected for one or more predictors (|z| > 10: {predictor_names}). This means the predictor perfectly (or near-perfectly) predict