HealthTowards Data Scienceabout 2 hours ago
Predictive Models Miss Causal Effects
1 min read
Variable selection for prediction drops confounders weakly tied to the outcome, biasing treatment effect estimates. Bayesian Adjustment for Confounding (BAC) forces exposure-model covariates into the outcome model. BIC-based selection and Bayesian model averaging both fail on finite samples.
Level
Hype check
Tap to vote and see what everyone thinks.
Summary by ByteBrief
