Thanks for the guidance. Clear on point 1.
Regarding point 2, what is the purpose of saving the values to a trace?
Regarding point 3, I understand the additive effect of the coefficients on the log-odds of the response variable. What I meant by my question is, if we specify pm.Deterministic, then PyMC will creates a separate coefficient for every row in the data.
(a). What does this coefficient mean in relation to the other variables (in this case the actor, block and treatment)? I suppose that if we’re looking at the coefficient for a particular row, then all the other coefficients don’t come into play.
(b). Why would we want this, it seems like overfitting?