國立政治大學統計學系
學 術 演 講
主講人：翁新傑助理教授(國立臺北大學統計系)學 術 演 講
題 目：Asymptotic Normality of Series Estimators for Semiparametric Mixed
Frequency Regression Models
時 間：民國114年12月8日 (星期一) 下午1：30
地 點：國立政治大學逸仙樓050101教室
摘 要：
This paper considers a class of semiparametric series estimator for the data sampled at different frequency, called MIxed DAta Sampling, or MIDAS, regression model. We focus on the performance of the widely known Almon lag polynomial for approximating the unknown weighting function of the higher frequency explanatory variables when the polynomial order is allowed to increase with the sample size of the dependent variable, because the resulting Almon estimator can be easily estimated with the OLS estimator and its asymptotic distribution is unexplored since it was first used for estimating the seminal autoregressive distributed lag (ADL) model in 1960s. Interestingly, as a dimension deduction approach, the Almon estimator for the long-run impact of the explanatory variable on the dependent one is shown to be asymptotically pivotal under suitable regularity conditions. Moreover, the Almon estimator is much more efficient than its unweighted counterpart as it can reduce the root of mean-squared-error (RMSE) of estimating the long-run impact by over 50%.
Keywords: Semiparametric regression, MIDAS model, Vandermonde matrix, Asymptotically pivotal.This is joint work with Wen-Jen Tsay.
