Double/debiased machine learning for logistic partially linear model

Summary We propose double/debiased machine learning approaches to infer a parametric component of logistic partially linear model. Our framework is based on Neyman orthogonal score equation consisting two nuisance models for the nonparametric model and conditional mean exposure with control group. To estimate models, we separately consider use high dimensional (HD) sparse regression (nonparametric) (ML) methods. In HD case, derive certain moment equations calibrate first order bias which preserves double robustness property. ML handle nonlinearity logit link through novel easy-to-implement ‘full refitting’ procedure. evaluate our methods simulation apply them in assessing effect emergency contraceptive pill early gestation new births 2008 policy reform Chile.