I would like to use a multivariate normal distribution with a diagonal covariance matrix in my model. What is the correct way to do it? At the moment I am using the following code: sigma = pm.InverseGamma('sigma', alpha, beta, shape=(num_dimensions)) mu = pm.MvNormal('mu', mu=means, cov=tt.nlinalg.…

Why not use the pm.Normal directly? mu = pm.Normal('mu', mu=means, sd = sigma, shape=(num_dimensions,))

Multivariate normal with diagonal covariance

junpenglao May 25, 2018, 11:22am 4

Yes they can be n-d tensor. We should update the docstrings if this point is not clear.

1 Like

Topic		Replies	Views
About set of Multivariate Normal Distributions Questions	2	1187	July 3, 2018
Generating a random matrix with MvNormal Questions	4	745	December 4, 2019
Model multivariate normal with separate means, dimension mismatch error Questions	2	3090	August 7, 2017
Assign cov in Likelihood of multi-variate Gaussian distribution Questions	4	847	February 5, 2018
Draw number of MvNormal in once Questions	2	575	April 5, 2019