Font Size: a A A

Model Detection For Varying Coefficient Panel Data Models With Fixed Effects

Posted on:2020-03-08Degree:MasterType:Thesis
Country:ChinaCandidate:X B AnFull Text:PDF
GTID:2417330575964016Subject:Statistics
Abstract/Summary:PDF Full Text Request
The varying coefficient models,which allow the coefficients in the model to be nonparametric functions,are very important non-parametric regression model,so they have better capacity of adaptability and modeling than the usual linear model;Meanwhile,the varying coefficient models still contain linear structure,so it is easy to be understood and explained in practice.The panel data combines the characteristics of time-series data and cross-section data.Compared with simple time-series data or cross-section data,the double dimensional panel data not only provide more individual information,improve estimation efficiency and control heterogeneity,but also can help researchers develop more complex modeling techniques.This paper study the model detection for panel data in varying coefficient models with fixed effects.Firstly,the unknown coefficient function in the model is expanded local linear method and the influence of fixed effect is eliminated by combining the dummy variable method.Secondly,based on adaptive LASSO method,a combined penalty estimation method is proposed to identify the real structure of the model by simultaneously penalizing coefficient function and their derivatives.That is to identify which covariates have functional effects on response variables,which variables have non-zero constant effects on response variables,and which variables have no significant effects on response variables.Under the appropriate regularity conditions,we prove the consistency,asymptotic normality and the consistency of model detection results.Finally,the performance of the combined penalty estimation method in finite samples is studied by numerical simulation.The organization of the paper is as follows,In section 1,we introduce the panel data,some correlation models,and the conventional method of selecting variables.We also introduce the current research results of the panel data and the main problems that we will solve in this paper.In section 2,we discuss the process of model detection in varying coefficient panel data model with fixed effects and the selection of the tuning parameters.In section 3,we give relevant conclusions about coefficient function estimation and model detection results.Numerical simulation and a real data analysis are reported in Section 4.The proofs of the main results are given in Section 5.
Keywords/Search Tags:Fixed effects, Panel data, Varying coefficient models, Variable selection, Oracle property, adaptive LASSO
PDF Full Text Request
Related items