If you are merely fitting a linear model, this is going to speed up things immensely. There are some good reasons for this, and I’ll leave it to you to check out ?priors to read more. Why? That’s because the priors for stan_lm puts a prior on it’s R 2. But, some details.įirst, you’ll note that we did not use stan_lm. There are defaults, largely following the recommendations of the STAN prior choice wiki. We’ll fit one poisson regression and one variable slope-intercept model as an example in each and look at inferential tools. While rethinking is awesome in the easy access it provides to stan, there are some simpler packages, and some more complex. Seriously, just check the CRAN Bayesian taskview (which isn’t even complete) (or look at python’s bayes packages) There is a whole ecosystem of tools designed to work with Bayesian data analysis. And likely to be updated any day now?īut what about software? We’ve been using the rethinking package which calls stan. I’ve also drawn very heavily on Gelman and Hill’s Data Analysis Using Regression and Multilevel/Hierarchical Models which is SUPER useful. Given what we’ve talked about, you should be able to translate to rethinking in a straightforward manner.
You’re now armed enough to be able to wrestle with Gelman et al’s Bayesian Data Analaysis vol 3 which is a fairly accessible intro to many many many models in the topic.
Coming to the end of this class, there are a few things you’ll find useful. 1.1 The wild world of Bayes beyond R class
0 kommentar(er)
