By Wayne
There are several reasons why everyone isn’t using Bayesian methods for regression modeling. One reason is that Bayesian modeling requires more thought: you need pesky things like priors, and you can’t assume that if a procedure runs without throwing an error that the answers are valid. A second reason is that MCMC sampling — the bedrock of practical Bayesian modeling — can be slow compared to closed-form or MLE procedures. A third reason is that existing Bayesian solutions have either been highly-specialized (and thus inflexible), or have required knowing how to use a generalized tool like BUGS, JAGS, or Stan. This third reason has recently been shattered in the R world by not one but two packages: brms and rstanarm. Interestingly, both of these packages are elegant front ends to Stan, via rstan and shinystan.
This article describes brms and rstanarm, how they help you, and how they differ.
You can install both packages from CRAN, making sure to install dependencies so you get rstan, Rcpp, and shinystan as well. If you like having the latest development versions — which may have a few bug fixes that the CRAN versions don’t yet have — you can use devtools to install them following instructions at the brms github site or the rstanarm github site.
The brms package
Let’s start with a quick multinomial logistic regression with the famous Iris dataset, using brms. You may want to skip the actual brm call, below, because it’s so slow (we’ll fix that in the next step):
library (brms) rstan_options (auto_write=TRUE) options (mc.cores=parallel::detectCores ()) # Run on multiple cores set.seed (3875) ir
First, note that the brm call looks like glm or other standard regression functions. Second, I advised you not to run the brm because on my couple-of-year-old Macbook Pro, it takes about 12 minutes to run. Why so long? …read more
Source:: r-bloggers.com