变系数固定效应面板数据模型的估计

Estimation of varying coefficient fixed effect models in panel data

在面板数据下基于变系数固定效应模型,采用两步估计方法和截面核估计方法来估计模型中的变系数部分及固定效应。模拟结果表明,当样本数据较小且观测次数较多时,两步估计法的估计效率略高于截面核估计的;而当样本数据较大且观测次数较少时,截面核估计方法的估计效率比两步估计的略好。本文通过两组不同样本的模拟数据验证了两种估计方法的有效性,说明变系数固定效应模型是处理面板数据的有效工具。

Two-step estimation and profile kernel estimation methods for the study of varying coefficient fixed effects models in panel data are proposed in this paper. Simulation shows that,for small samples with a large number of observations,the two-step estimation method is more efficient than profile kernel estimation. In contrast,for large samples with few observations,the accuracy of profile kernel estimation is better than two-step estimation. Both methods of estimation are efficient and the varying coefficient fixed effect models are powerful tools for panel data analysis.