A Class of Semiparametric Models with Homogeneous Structure for Panel Data Analysis
share
Stimulated by the analysis of a dataset from China about Covid-19, we propose a class of semiparametric models for panel data analysis. The proposed models account for both homogeneity and heterogeneity among the individuals of a panel data. They strike a nice balance between parsimony and risk of misspecification. Although stimulated by the analysis of a particular dataset, the proposed models apply to very broad range of panel data analysis, they are powerful in exploring nonlinear dynamic patterns of impacts of covariates or transformed covariates. An estimation procedure is presented, and its asymptotic properties are established. Intensive simulation studies are also conducted to demonstrate how well the estimation procedure works and the risk of ignoring homogeneity or heterogeneity among individuals in panel data analysis. Finally, we apply the proposed models and estimation procedure to the Covid-19 data from China, and reveal some interesting dynamic patterns of the impacts of some important factors.
READ FULL TEXT