Localising Dropout Variance in Twin Networks

Accurate individual treatment-effect estimation demands not only reliable point predictions but also uncertainty measures that help practitioners locate the source of model failure. We introduce a layer-wise variance decomposition for deep twin-network models: by toggling Monte Carlo Dropout independently in the shared encoder and the outcome heads, we split total predictive variance into an encoder component (σenc2) and a head component (σhead2), with σenc2 + σhead2 ≈ σtot2 by the law of total variance. Across three synthetic covariate-shift regimes, the encoder component dominates under distributional shift (ρenc=0.53) while the head component becomes informative only once encoder uncertainty is controlled. On a real-world twins cohort with induced multivariate shift, only σenc2 spikes on out-of-distribution samples and becomes the primary error predictor (ρenc!≈!0.89), while σhead2 remains flat. The decomposition adds negligible cost over standard MC Dropout and provides a practical diagnostic for deciding whether to collect more diverse covariates or more outcome data.