# Setting Different Priors on Category-Specific Coefficients

**URL:** <https://discourse.mc-stan.org/t/setting-different-priors-on-category-specific-coefficients/7565>\
**Category:** brms\
**Created:** [February 1, 2019, 9:06pm UTC](https://discourse.mc-stan.org/t/setting-different-priors-on-category-specific-coefficients/7565 "2019-02-01T21:06:45Z")\
**Posts on this page:** 2\
**Page:** 1

<div class="post-metadata">

**Author:** ![rpr2151](https://avatars.discourse-cdn.com/v4/letter/r/85f322/32.png) [@rpr2151](https://discourse.mc-stan.org/u/rpr2151)\
**Post date:** [February 1, 2019, 9:06pm UTC](https://discourse.mc-stan.org/t/setting-different-priors-on-category-specific-coefficients/7565/1 "2019-02-01T21:06:45Z")

</div>

I’m fitting an adjacent categories ordinal model with category-specific effects, and I’d like to set different priors on the per-level coefficients. For a reproducible example, consider the following model:

library(brms)  
library(foreign)  
dat \<- read.dta(“[https://stats.idre.ucla.edu/stat/data/ologit.dta](https://stats.idre.ucla.edu/stat/data/ologit.dta)”)  
dat$apply = as.integer(dat$apply)  
fit1 = brm(bf(apply~cs(gpa)), family = acat(), data = dat, chains = 2)

GPA has two coefficients, as expected, but prior\_summary(fit1) shows only one possible prior location for GPA, which sets the priors for both of the coefficients. How would I go about setting, for example, a uniform(-0.001,0.001) prior on gpa[1] (essentially setting it to 0), while leaving gpa[2] with more traditional prior?

- Operating System: Ubuntu 16.04
- brms Version: 2.7.0

---

<div class="post-metadata">

**Author:** ![paul.buerkner](https://yyz2.discourse-cdn.com/flex030/user_avatar/discourse.mc-stan.org/paul.buerkner/32/3303_2.png) [@paul.buerkner](https://discourse.mc-stan.org/u/paul.buerkner)\
**Post date:** [February 4, 2019, 11:06am UTC](https://discourse.mc-stan.org/t/setting-different-priors-on-category-specific-coefficients/7565/2 "2019-02-04T11:06:16Z")

</div>

Currently, brms does not support varying priors over category specific effects, but please feel free to open an issue on [https://github.com/paul-buerkner/brms](https://github.com/paul-buerkner/brms) so that I may eventually implement it.