I found problems with the estimation of the Gaussian copula (standalone or in a mixture of copulas).
The main issue is the usage of the scipy.stats.multivariate_normal module in the calculation of the pdf of Gaussian copula.
As such I avoided it and calculated the Gaussian copula density explicitly. This change fixed the estimation issues.
I found problems with the estimation of the Gaussian copula (standalone or in a mixture of copulas). The main issue is the usage of the
scipy.stats.multivariate_normalmodule in the calculation of the pdf of Gaussian copula. As such I avoided it and calculated the Gaussian copula density explicitly. This change fixed the estimation issues.