摘要: Various special linear structures connected with covariance matrices are reviewed and graphical methods for their representation introduced, involving in particular two different kinds of edge between the nodes representing component variables. The distinction decomposable nondecomposable is emphasized. Empirical examples described main possibilities four