关键词: 一致模型选择, 门限因子模型, 收缩估计

Abstract: With the threshold-type regime shift in factor loadings, the threshold factor model can delineate the co-movement and regime shift of high-dimensional time series. In this paper, a consistent model selection process of high-dimensional threshold factor model is presented based on an adaptive group-LASSO method, in which the identification of the number of factors and the inference of the threshold effect are integrated into a unified analytical framework. It not only solves the consistency problem of model selection, but also realizes the unified control of model selection error, which is very important for a high-dimensional threshold factor model. The results of theoretical research and random simulation show that the method has a good large sample property and limited sample performance. Finally, the threshold factor model is applied to China’ s financial market, and the empirical results further verify the effectiveness of our theory.

Key words: Consistent Model Selection, Threshold Factor Model, Shrinkage Estimation