Model averaging with high-dimensional dependent data

The past two decades witnessed a prosperous literature on model averaging, however, few authors have examined model averaging under high-dimensional data setting. An exception is Ando and Li (2014), which proposed a model averaging procedure to improve prediction accuracy under high-dimensional independent data setting. In this paper, we broaden Ando and Li's scope of analysis to allow dependent data. We show that under the dependent data setting, their model averaging estimator is still asymptotically optimal. Simulation study demonstrates the finite sample performance of the estimator in a variety of dependent data settings.