Prior and sampling

Algorithms
1

Is the prior for a parameter, when specified, also used for sampling? Or only to update the likelihood?
I’m asking this as specifying a uniform prior, as follows:

 l~uniform(0.1,20)

and printing on the screen the sample values during the runtime, I obtained some samples that fell outside the interval [0.1,20].
This, however, didn’t happen when I specified a lower and an upper bound.

My hypothesis is that the lower and the upper bounds constrain the samples to belong to an interval (and hence, samples are drawn from the uniform distribution), while the prior is used only to “regularise” the likelihood, but I’m not sure of that.

Thanks for the help

Cesare

2

That is true and in the documentation. The prior distributions used in the model block provide a topology for that subspace.