Generalized Empirical Likelihood Inference in Semiparametric Regression Model for Longitudinal Data

<正> In this paper,we consider the semiparametric regression model for longitudinal data.Dueto the correlation within groups,a generalized empirical log-likelihood ratio statistic for the unknownparameters in the model is suggested by introducing the working covariance matrix.It is proved thatthe proposed statistic is asymptotically standard chi-squared under some suitable conditions,and henceit can be used to construct the confidence regions of the parameters.A simulation study is conducted tocompare the proposed method with the generalized least squares method in terms of coverage accuracyand average lengths of the confidence intervals.