Combine partial covariance matrices using a Wishart-EM algorithm. Methods are described in the November 2019 article by Akdemir et al. < https://www.biorxiv.org/content/10.1101/857425v1>. It can be used to combine partially overlapping covariance matrices from independent trials, partially overlapping multi-view relationship data from genomic experiments, partially overlapping Gaussian graphs described by their covariance structures. High dimensional covariance estimation, multi-view data integration. high dimensional covariance graph estimation.