Thomas Lumley • 10/18/2015

Double robustness

The article delves into the statistical methodology of 'double robust' estimation in regression. It clarifies that this approach uses one model for the outcome (Y) and another for the exposure (X), given available data (Z). The estimates are consistent if either model is correct and efficient if both are correct. It argues the technique is more useful than it initially sounds, as bias terms multiply and different substantive knowledge can inform each model.

0 comments

#Causal Inference #Statistical Inference #Regression Analysis