Can I make out-of-sample predictions using posterior means? Or how to use scipy.optimize with pymc predictions?

[ricardoV94]

I usually just use a thinned version of the posterior and then take the mean of all the losses