Local Information Theoretic Methods for smooth Coefficients Dynamic Panel Data Models
本文提出一类局部估计量用于半参数平滑系数动态面板数据模型，该估计量具有信息论解释，并可用于检验模型设定和系数恒定性，适用于不同N和T大小以及变量相关或不可观测的情况。
This paper considers estimation and inference in semiparametric smooth coefficients dynamic panel data models. It proposes a class of local estimators that can be given an interesting information‐theoretic interpretation and a number of test statistics that can be used to test for the (local) correct specification of the model and for the constancy of the smooth coefficients. The results of the paper are rather general as they allow for the three cases of ‘large N , small T ’, ‘small N , large T ’ and ‘large N , large T ’, for the possibility that some of the regressors might be correlated with the unobservable errors and for the possibility that some of the variables used in the estimation might not be directly observable. Simulations show that the proposed method have competitive finite sample properties.