Modal regression for fixed effects panel data

Most research on panel data focuses mean or quantile regression, while there is not much about regression methods based the mode. In this paper, we propose a new model named fixed effects modal for in which how conditional mode of response variable depends covariates and employ kernel-based objective function to simplify computation. The proposed can complement regressions provide better central tendency measure prediction performance when are skewed. We present linear dummy method pseudo-demodeing two-step estimate regression. computations be easily implemented using modified modal–expectation–maximization algorithm. investigate asymptotic properties estimators under some mild regularity conditions number individuals, N, time periods, T, go infinity. optimal bandwidths with order $$(NT)^{-1/7}$$ obtained by minimizing weighted squared errors. Monte Carlo simulations two real analyses public capital productivity study carbon dioxide emissions presented demonstrate finite sample newly