Abstract:

The authors concerned  the partial linear models with serially correlated random errors which are not observed and modeled by a moving\|average process of infinte order. We proposed an empirical loglikelihood ratio statistic for the regression coefficients. Our results show that the statistic is asymptotically chisquare distributed and the corresponding confidence interval can be constructed accordingly. A simulation illustrates that the empirical likelihood method works better than the ordinary least squares method.

Key words: partial linear regression model, MA(∞) error process, empirical likelihood