交互效应面板模型的EM算法和MCMC算法

摘要

The panel data model with interactive effects is one of active research fields on the frontier of econometrics, which have wide applications in empirical re- search. However, how to get the estimators of the parameters in the model is a big challenge to many empirical economists. The usual Newton-Raphson algorithm is not a good choice since the optimization frequently fails for the special structure of the factor models. To accommodate this issue, this paper designs the EM algorithm and MCMC algorithm to deal with this problem. The experiment over the computer indicates that the both algorithms work very well in the finite sample.%交互效应面板模型是目前计量经济学前沿研究的热点，有着广阔的应用空间。但是对很多应用者而言，模型内的参数估计是一个非常棘手的问题。通常的Newton-Raphson算法在优化似然函数的过程中，常常会出现优化失败的情况。本文依据EM算法和MCMC算法理论，为应用研究者提供了一套获得参数估计值的流程。计算机上的试验证实两种估计方法都非常稳健可靠，并在很多情况下，差异不是很大。