Checking understanding of bounds on priors

Interfaces brms
1

I just wanted to confirm whether or not I understand setting upper and lower bounds on priors, for the types of parameters where this is possible.

If I set a prior on a parameter, e.g., normal with mean 50 and sd of 20, and then set the lb and ub to 0 and 100 respectively, I just want to confirm that this means the prior will just be from 0-100. As a more extreme example, if I set a prior as normal with mean 50 and sd 50 but bound it from 0-100, this cuts off about 15-16% of the distribution on either side. Would this then reflect a prior that is peaked around 50, views values close to 0 or 100 as not highly improbable, and does not consider values below 0 or above 100? Probabilities within the cutoffs are then made to sum to 1, and there are cliffs in the probability for values below 0 and above 100.

Thanks for your help.

Please also provide the following information in addition to your question:

  • Operating System: MacOS Sierra 10.12.6
  • brms Version: 2.8.0
2

Yes exactly. By using lb and ub your are effectively created a truncated prior with support only between lb and ub.

1 Like
3

Perfect, thank you!