Trying to increase my practical knowledge of adjusting my models to accommodate different distributions. Any suggestions on good resources on how to further probe + remedy potential concerns highlighted by posterior predictive checks? #Rstats#StatsSky

Mixture of gammas or log normals

In your case, it appears you're trying to model non-negative continuous data, such as response times, with the Gaussian likelihood. Reaction-time data have long been described as approximately lognormal. The gamma distribution is another fine alternative (as discussed by McElreath).

Statistical Rethinking, chapter 10. Here's my bookdown rendition: bookdown.org/content/4857...