Matlab copulastat and copularnd equivalent about copulas HOT 2 CLOSED

Dear Manuel
thanks, actually googling a bit I was able to find out almost the same,
and indeed I already implemented it some dais ago it in my to be cleaned
minimal https://github.com/lstorchi/markovctheil/blob/master/copula_cli.py
copula CLI .

thanks again

from copulas.

Hi @lstorchi, and thanks for your questions,

I'm not Matlab expert, and the answers are based on what I've been able to grasp after reading the documentation of copulastat and copularnd modules, so please correct me if I'm wrong

r = copulastat('Gaussian',rho) returns the Kendall’s rank correlation, r, that corresponds to a Gaussian copula with linear correlation parameters rho.

If I'm not mistaken, what you intend to do, can't be done directly using our package, but can be done easily using pandas.DataFrame.corr directly on your raw data.

u = copularnd('Gaussian',rho,n) returns n random vectors generated from a Gaussian copula with linear correlation parameters rho.

We have not implemented a way to sample from a copula passing directly the linear correlation parameters. What we do is compute the correlation from the raw data, and when sampling, we return the sampled values in the same space that the raw data.

However, as the gaussian copula is based on the Gaussian Multivariate distribution, you can archieve this by doing:

import numpy as np

# We are assuming `rho` is your correlation parameters in a numpy.array
# We are assuming `n` is the number of vector you want to sample.

size = rho.shape[0]
means = np.zeros(size)
np.random.multivariate_normal(means, rho, size=n)

from copulas.