Generating orthogonal latent variables

10

Thanks for that info Bob. It looks like in that julia forum answer I linked which is elaborated on a bit in this paper https://arxiv.org/pdf/math-ph/0609050 they handle the determinant issue by multiplying Q by the sign of the diagonal of R. I gave this a try but pymc says “ValueError: Model can not be sampled with NUTS alone. Your model is probably not continuous.” which I guess is due to the sign function. In any case, I think I will stick with lkj with a large eta for now since this does not seem to have any issue sampling with nuts. Thanks all!