| In this paper,the statistical inference problem of two kinds of regression models under longitudinal data is studied by using composite quantile regression estimation method.Random effects model is one of the most commonly used model in longitudinal data processing.For the random effects model,using the composite quantile regression estimation to estimate the parameters of the model,and in some regular conditions show that the estimated asymptotic normality.Through simulation research,compared with the traditional least squares estimation,the median regression estimation and composite quantile regression estimate the accuracy of the simulation results show that in the case of limited samples,this paper proposed the method of random effects model parameter estimation is effective,especially when the error term of model don't obey the normal distribution,The Bias,SD and RMSE of compound quantile regression estimation were all optimal.In the selection of the quantile,K= 9 is simple and accurate.Finally,the method proposed in this paper was applied to the data of blood alcohol content after drinking,and good results were obtained.Variable coefficient model is an extension of the linear model,it has better flexibility and explanatory,especially suitable for the analysis of longitudinal data.For variable coefficient model,in a composite quantile regression framework,the objective function is constructed based on local polynomial approximation,to solve model of objective function coefficient function and variance function estimates,drawing the large sample properties of the estimates under the condition of a certain regular.In the study,using the weighted least squares estimation and composite quantile regression estimation,the simulation results show that in the case of limited samples,the method proposed in this paper is effective for estimating variable coefficient models with different variance structures and performs well under different variance structures. |
PDF Full Text Request