where to put prior_alpha with a beta prior

[dirknbr]

If I have a beta prior where do I put the prior_alpha

target += beta_lpdf(e | prior_alpha, 1);

or

target += prior_alpha * beta_lpdf(e | 1, 1);

[n-kall]

Regardless of the prior, the second option is the way to do it.

But this is only necessary if the pareto smoothed importance sampling estimates are not reliable, and you want to refit the model with different alpha values.

[dirknbr]

ok, so the second option will not make my e go outside the (0, 1) range?

additional question, if we have 2 parameters and use one prior_alpha, do we measure the joint effect effect of a changed prior_alpha?

[n-kall]

The support won't change because the alpha is changing the density values, not the parameter values themselves. For details and example, you can check the paper linked in the readme.

If you have just one prior_alpha and multiply with multiple log prior densities, you'll be checking the effect of power scaling all those priors at the same time