Fits a logistic regression using a single covariate (typically age or sex)
as the only predictor of case status. Quantifies how much of the apparent
classifier signal could be explained by demographic confounding alone.
Usage
os_covariate_only_auc(y, covariate, covariate_name = "covariate")
Arguments
- y
Binary outcome vector.
- covariate
Numeric or factor covariate (e.g., age in years).
- covariate_name
Human-readable label for reporting.
Value
A list:
- auc
AUC of covariate-only classifier
- direction
"higher values = cases" or "higher values = controls"
- effect_size
Cohen's d (numeric) or Cramer's V (factor)