Abstract:
The modal factor model represents a new factor model for dimension reduction in high dimensional panel data. Unlike the approximate factor model that targets for the mean factors, it captures factors that influence the conditional mode of the distribution of the observables. Statistical inference is developed with the aid of mode estimation, where the modal factors and the loadings are estimated through maximizing a kernel-type objective function. For practical implementation, an alternating maximization algorithm is designed to obtain the estimators. Two model selection criteria based on information criteria and rank estimation are also proposed to determine the number of factors. The consistency of the proposed estimators and the asymptotic normality of the modal factor estimators are established under some regularity conditions. Simulations demonstrate the nice finite sample performance of our proposed estimators, even in the presence of heavy-tailed and asymmetric idiosyncratic error distributions. Empirical applications illustrate the practical merits of modal factors in forecasting the U.S. inflation rate and real GDP growth.
Speaker's Personal Website: https://en.gsm.pku.edu.cn/faculty/yundong.tu/
