We should implement a variant of PLN for structuring the covariance matrix on top of 'diagonal', "spherical" and "full", with the following form:  where C is a positive definite given correlation matrix, and sigma and nu are to be estimated