Title: A model-averaging approach for functional-coefficient regression
Authors: Yuying Sun - Academy of Mathematics and Systems Science, Chinese Academy of Sciences (China) [presenting]
Zongwu Cai - The University of Kansas (United States)
Shouyang Wang - Academy of Mathematics and System Science, Chinese Academy of Sciences, (China)
Abstract: Model averaging aims at providing an insurance against selecting a poor forecast model. All existing model averaging approaches in the literature are designed with constant combination weights. Little attention has been paid to functional weighing in model averaging, which is more realistic in economics and finance. A novel model averaging estimator is proposed which selects optimal functional combination weights by minimizing a local leave-subject-out cross-validation criterion. It is shown that the proposed functional leave-subject-out cross-validation model averaging (FLsoMA) estimator is asymptotically optimal in the sense of achieving the lowest possible local squared error loss in a class of functional model averaging estimators. Under a set of regularity assumptions, the FLsoMA estimator is root-Th consistent. A simulation study and an empirical application highlight the merits of the proposed FLsoMA estimator relative to a variety of popular estimators with constant model averaging weights and model selection.