`stack`ing and `sum`ing a list of random variables ruins sampling?

Thanks for your help - yes, I could probably rewrite that as a .dot of a parameter tensor with all my data columns.

I was under the impression that the pm.math.stack would somehow transform my list of 1D tensors into a 2D tensor (so that it wouldn’t matter any more that it had been a list). But apparently it’s not that easy.

Thanks a lot!