KBFE: A Knowledge Based Fixed Effect Model for Nonlinear Constraints

Abstract Fixed effects are widely recognized as powerful econometric methods for heterogeneity analysis in panel data sets. Such as variable selection, parameter estimation and dimension reduction has been taken into account on linear assumptions in the existing research. However, the nonlinear constraints are just taken into account for Boolean check but not digital analysis. In this paper, a knowledge based fixed effect model has been proposed for nonlinear constraints with effect factors. Firstly, parameter has been well estimated following the expert knowledge for the nonlinear fitting in an abstract way. Secondly, the co-effect, time effect, and individual effects have been calculated by function separation. Lastly, the effectiveness of the knowledge based fixed effect model has been compared with the nonlinear method on a synthetic data.