Brms paper published

Paul’s paper on the brms R package is officially out at the Journal of Statistical Software

https://www.jstatsoft.org/article/view/v080i01

For those of you who don’t know what it does, it has one primary R function called brm that allows you to specify many, many regression models using (a somewhat expanded version of) the familiar formula, data.frame, and additional arguments syntax that has been used to specify models in R or S for decades. It then generates the corresponding Stan code at runtime and passes it off to the functions in the rstan package to compile and execute the Stan program. After the results come back, you can use bayesplot, loo, and shinyStan packages in much the same way as with rstanarm.

I continue to be puzzled by how many people start writing a Stan program with a blank text file. It is much less error prone to call brms::make_stancode to generate a working, commented Stan program that can be edited into the one you want to estimate and then calling brms::make_standata to create a named list of data to pass to Stan.

@article{JSSv080i01,
author = {Paul-Christian Bürkner},
title = {brms: An R Package for Bayesian Multilevel Models Using Stan},
journal = {Journal of Statistical Software, Articles},
volume = {80},
number = {1},
year = {2017},
keywords = {Bayesian inference; multilevel model; ordinal data; MCMC; Stan; R},
abstract = {The brms package implements Bayesian multilevel models in R using the probabilistic programming language Stan. A wide range of distributions and link functions are supported, allowing users to fit - among others - linear, robust linear, binomial, Poisson, survival, ordinal, zero-inflated, hurdle, and even non-linear models all in a multilevel context. Further modeling options include autocorrelation of the response variable, user defined covariance structures, censored data, as well as meta-analytic standard errors. Prior specifications are flexible and explicitly encourage users to apply prior distributions that actually reflect their beliefs. In addition, model fit can easily be assessed and compared with the Watanabe-Akaike information criterion and leave-one-out cross-validation.},
issn = {1548-7660},
pages = {1–28},
doi = {10.18637/jss.v080.i01},
url = {https://www.jstatsoft.org/v080/i01}
}

6 Likes

Brms is fantastic. It’s a shame it took so long for this to be published, given that brms does much more than the paper suggests at this point.

I blame the reviewers.

1 Like