Econometrics Seminar -Thomas Tao Yang (ANU)
Title: High Dimensional Binary Choice Model with Unknown Heteroskedasticity or Instrumental Variables With Fu Ouyang (University of Queensland)
Abstract: In this paper, we consider binary choice model estimation in a high dimensional setting. We do not assume any distributional assumption on the error term, and we permit the error to be heteroskedastic and correlated with regressors. The approach we propose is a modified special regressor method. The original special regressor estimator, proposed by Lewbel (2000), is generally not feasible in the high dimensional setting because it involves high dimensional conditional density estimation. We propose a novel method to reduce the dimension of and estimate the conditional density. With the new feasible estimator, we study the variables and instrument variables selections for the high dimensional binary choice model.