Estimation Of Semi-parametric Varying Coefficient Panel Data With Measurement Error Models

Semi-parametric varying coefficient Panel data models combine with the characteristics of the Panel model,and the semi-parametric varying coefficient models.It is an effective tool for dealing with the data.In real life,often encounter with measurement error data.Therefore,In this paper,the estimation problem of the semi-parametric variable coefficient Panel model under measurement error data is studied.The main work is as follows:1.For the heteroscedastic semi-parametric varying coefficient Panel data models with measurement error of associated explanatory variables of parametric component and random effects.The problem of estimating parametric and nonparametric components in this models are studied.Assume that the random error term is a homogeneous variance structure,the first estimates constructed by the modified profile least squares method.Based on this estimate,the variance of the individual effects and a kernel estimate of variance function is obtained,which in turn is used to define re-weighted estimates of the parametric and varying coefficients function of the semi-parametric varying coefficient Panel model with measurement error models.Under certain conditions,proved the large sample property of the estimate.The finite sample performance of the proposed estimates are evaluated via simulation studies.The results show that re-weighted estimation is effective.2.For the partially linear time-varying coefficient Panel data models with measurement error of associated explanatory variables of parametric component and fixed effects.The problem of estimating parametric and nonparametric components in this models are studied.Due to the identifiability of the model,so that the data can be carried out within the group average.Then the model is transformed into a partially time-varying coefficient EV model,the parameters and the non parametric part in the model are estimated by modified least squares profile method.Under the some regular conditions,the asymptotic properties of the estimation are obtained.Some simulation studies are conducted to examine the finite sample performance for the proposed method,which considered the correlated with the fixed effects and the regression.