Propensity score estimation: Comparison of logistic regression, deep neural network, and convolutional neural network
This study proposes two machine learning techniques—deep neural network (DNN) and convolutional neural network (ConvNet)—as new estimators of propensity scores that can algorithmically handle nonlinear relationships and interactions of covariates. A simulation was conducted to examine the performance of DNN, Convnet, and logistic regression in comparison to the conventional method. In terms of bias reduction, the performances of propensity scores estimated by DNN and Convnet were better than the performance of one that was estimated by logistic regression overall in most cases. Therefore, this study confirmed the feasibility of DNN and ConvNet as propensity score estimators and provides guidelines to use them.Embargo status: Restricted until 09/2172. To request the author grant access, click on the PDF link to the left.