blme icon indicating copy to clipboard operation
blme copied to clipboard

list for fixef.prior?

Open snuggleduck opened this issue 1 year ago • 1 comments

Hi Vincent, I'm trying to implement the method of Abrahantes & Aerts (2012; https://doi.org/10.1177/1471082X1001200102) with bglmer in a binomial model with more than one fixed effect, so I would need to center and scale a Cauchy prior around a different value for each fixed coeffient. Or I would even want to have different distribution families for different fixed coefficients (e.g. Cauchy for coefficients with separation problem and normal flat for coefs that have no separation issue). But as far as I can see from the reference manual, fixef.prior will only accept a single distribution that is then applied to all fixed coefficients. Would it be possible to formulate fixef.prior like cov.prior, as a list of as many distributions as coefficients? Thanks in advance for any help. Cheers, Veronika

snuggleduck avatar Jan 17 '25 21:01 snuggleduck

It's not currently implemented but it wouldn't be too hard to extend the existing t distribution implementation to treat each coordinate as independent and allow for a different degree of freedom in each. I don't know that I'll necessarily have time in the near future but if I do I'll upload something here.

vdorie avatar Jan 22 '25 00:01 vdorie