固定交互效应面板截取模型的ECM算法

ECM Algorithm for Panel Censored Model with Interactive Fixed Effects

交互效应面板数据模型在社会经济问题的实证分析中具有很强的适用性,但现有研究主要集中于线性面板模型。本文将交互效应引入非线性的面板截取模型,并基于ECM算法,建立了有效估计量和识别程序。基于不同因子类型的仿真实验结果显示,ECM算法可以很好地识别面板截取样本中的非观测因子。ECM估计量具有良好的有限样本性质,与其他估计量相比具有更小的偏误和更快的收敛速度,尤其是当共同因子为低频平滑因子时,其表现最为理想。

Panel data models with interactive effect can reveal the different effect of common factor on individuals. In this paper, we consider a panel censored model with interactive effect and propose ECM algorithm for censored factor analy- sis of censored sample and conditional maximum likelihood estimation. A simula- tion study shows that unobserved factors in censored panel data can be identified well By ECM algorithm, and ECM estimator has perfect finite sample properties since it exhibits the minimum bias compared with other alternative estimators espe- cially when the common factor is smooth. Also, the bias dies out quickly as the sample size increases.