Semiparametric time series regression modeling with a diverging number of parameters

摘要

Variable selection and error structure determination of a partially linearrnmodel with time series errors are important issues. In this paper, werninvestigate the regression coefficient and autoregressive order shrinkagernand selection via the smoothly clipped absolute deviation penaltyrnfor a partially linear model with a divergent number of covariatesrnand finite order autoregressive time series errors. Both consistencyrnand asymptotic normality of the proposed penalized estimators arernderived. The oracle property of the resultant estimators is proved.rnSimulation studies are carried out to assess the finite-sample performancernof the proposed procedure. A real data analysis is made tornillustrate the usefulness of the proposed procedure as well.