Thomas Lumley • 10/25/2014

Semiparametric efficiency and nearly-true models

This technical article discusses statistical efficiency in two-phase sampling, where additional variables are measured on a subset of a larger cohort. It compares two semiparametric models (Model D and Model M) for estimation, focusing on scenarios where the parametric model is 'nearly true' but slightly misspecified. It explores the trade-offs between bias and precision using concepts like contiguity and references estimators like AIPW and unweighted logistic regression.

0 comments

#Logistic Regression #Semiparametric Models #Statistical Efficiency