Model Averaging for Count-Data Autoregressive Models with Covariates

Authors: Jiajing Sun, Zhuo Lin and Brendan McCabe.

Original title: 带协变量的计数数据自回归模型的模型平均

When the correct model for a series of counts is uncertain, forecasts can combine several candidate models. This paper develops a K-fold cross-validation model-averaging method for integer-valued autoregressive processes with covariates, fitted by maximum likelihood. It evaluates point forecasts, intervals and entire predictive distributions. The authors establish asymptotic optimality under squared loss and study forecast accuracy through simulations and an empirical application.

Recommended citation: Sun, Jiajing, Zhuo Lin, and Brendan McCabe. “带协变量的计数数据自回归模型的模型平均” [Model Averaging for Count-Data Autoregressive Models with Covariates]. China Journal of Econometrics. Accepted for publication. In Chinese.