I propose that that we should have a function on multivariate distribitions that marginalizes out variates/
For example going from a 3D Multivariate to a 2D one.
Marginalizing is probably not trivially defined on all multivariate distributions.
But it is defined on MvNormal and GenericMvTDist
where you can basically just drop the relevent rows/columns from the matrixes that live behind it.
Invenia currently has some internal code for this that i can open source.
I propose that that we should have a function on multivariate distribitions that marginalizes out variates/ For example going from a 3D Multivariate to a 2D one.
Marginalizing is probably not trivially defined on all multivariate distributions. But it is defined on
MvNormal
andGenericMvTDist
where you can basically just drop the relevent rows/columns from the matrixes that live behind it. Invenia currently has some internal code for this that i can open source.