Hi my name is Andy. I am new to brms and Bayesian analysis in general.

I’m trying to understand hurdle models and playing around with understanding effects of priors on hurdle_lognormal with prior predictive checks.

My question is more basic. I noticed with prior_summary(), the default prior for Intercept_hu is “logistic(0, 1)”, rather than a distribution. What does that mean? The only other discussion I could find was here: Choosing priors for a brms hurdle model where jobi asked if putting a prior as “logistic(3, 1)” was appropriate. What does “logistic(3, 1)” as a prior mean? Would a normal(0, 1.5) be appropriate as a prior for the intercept in hurdle models?

I couldn’t find much about priors for hurdle models on google or stack exchange either, so am wondering if anyone can point me to the right direction.

Thank you very much.

1 Like

I can’t help you with hurdle models but I will point out that there is a logistic distribution. It looks roughly like a normal distribution but with fatter tails. The logistic distribution is parameterized with a location \mu and a scale s. I assume, not having tried to find the function definition, that logistic(0,1) has \mu = 0 and s = 1. I plotted logistic(0,1), using the dlogis function in R, and normal(0,1.5) and they are fairly similar, though the logistic distribution does have a little more amplitude beyond roughly |2.5|.

1 Like

Thank you so much! I knew it was something basic and relating to my lack of knowledge! Thanks again!