具有不可忽略退出的面板数据混合半参数变系数模型

Mixtures of semiparametric varying coefficient models of panel data with nonignorable dropout

由于面板数据经常出现由信息性退出而引起的缺失数据,故其统计分析比较复杂。本文假定响应变量服从部分线性半参变系数混合效应模型,其中非参数系数函数依赖于相应的退出时间。利用二步估计方法的思想求得参数与非参数部分的相合估计,减少了同时估计的参数的个数,避免了估计方程中由较多参数引起的多重共线性问题。

The analysis of panel data is frequently complicated by missing data due to information dropout. We assume that responses follow a semiparametric varying coefficient random effect model. The concept of two step estimation is used in order to provide consistent estimators of the parameteric and nonparametric parts. The methods simplify the calculations and avoid the collinearity caused by more parameters in the model.