Factorized estimation of high?dimensional nonparametric covariance models

Estimation of covariate-dependent conditional covariance matrix in a high-dimensional space poses challenge to contemporary statistical research. The existing kernel estimators may not be locally adaptive due using single bandwidth explore the smoothness all entries target function. Moreover, corresponding theory holds only for i.i.d. samples although most applications, are dependent. In this paper, we propose novel estimation scheme overcome these obstacles by techniques factorization, thresholding and optimal shrinkage. Under certain regularity conditions, show that proposed estimator is consistent with underlying even when sample We conduct set simulation studies significantly outperforms its competitors. apply procedure analysis an asset return dataset, identifying number interesting volatility co-volatility patterns across different time periods.