A1258
Title: Doubly robust estimation of average partial effects
Authors: Harvey Klyne - University of Cambridge (United Kingdom) [presenting]
Rajen D Shah - University of Cambridge (United Kingdom)
Abstract: Single-parameter summaries of variable effects are desirable for ease of interpretation, but linear modelling assumptions commonly result in poor model fitting. A modern approach is to estimate the average partial effect---the average slope of the regression function with respect to the predictor of interest---using a doubly robust semiparametric procedure. Existing work has focused on specific forms of nuisance function estimators. The scope is extended to arbitrary plug-in nuisance function estimation, allowing for the use of modern machine learning methods. The procedure involves resmoothing a first-stage regression estimator to produce a differentiable version and modelling the joint distribution of the predictors through a location-scale model. It is proven that the proposals lead to a semiparametric efficient estimator under weak assumptions, and attractive numerical performance is demonstrated even under misspecification.
