Announcement: 'shinybrms'

I don’t know if The Stan Forums are the right place for this announcement, but I thought it might not harm:

Recently, my R package shinybrms was released on CRAN (see here). The corresponding GitHub page may be found here.

This package provides a GUI (a Shiny app) for fitting Bayesian regression models using brms. Of course, most of you don’t need a GUI for fitting Bayesian regression models. But shinybrms was developed to spread Bayesian regression models out to a broader public which is not familiar with R. Perhaps some of you are giving advise in Bayesian regression modeling to people from this broader public and want to give those people the opportunity to fit their models themselves. That’s where shinybrms comes in. Since it relies on Shiny, shinybrms may be run on a server and then accessed via a web browser (so that users don’t even need to install R and the necessary R packages).

The current version is 1.0.1 and offers only limited features:

  • For the (univariate) outcome: only a Gaussian, Bernoulli, or negative binomial distribution.
  • For the predictors: Only non-varying (a.k.a. population-level or “fixed”) effects are supported. Varying (a.k.a. group-level or “random”) effects are not supported yet. Interactions are supported, though.
  • Most of brms's special features, like monotonic effects for ordinal predictors, non-linear effects, or the possibility to specify standard errors for the outcome (for meta-analyses and meta-regressions) are not supported yet.
  • For the inspection of the output, only a short summary (from brms::summary.brmsfit()) and the possibility to launch shinystan is offered.

Of course, the missing features are on my to-do list. Additionally, I plan to include (for example):

  • a custom output inspection directly in shinybrms (additionally to the possibility to launch shinystan),
  • model selection possibilities.

Further suggestions for improvements are very welcome.

Note that the current version has not been tested yet with R 4.0.0 (as shinybrms v1.0.1 has been released on CRAN on March 26, 2020, i.e. before the final release of R 4.0.0). This is the first thing on my to-do list.

10 Likes

As good as any! Certainly welcome

@andrewgelman, @mitzimorris have a look at this (it fits in the theme of Stan via the interwebs).

A few days ago, version 1.1.0 has been published on CRAN (see here). Varying effects are supported now. See the NEWS for details.

5 Likes

Cool work. Could you add couple screen shots, for example, to github pages?

1 Like

Sure, good idea.

1 Like

Screenshots are now available at https://fweber144.github.io/shinybrms/articles/shinybrms.html.

5 Likes

Recently, shinybrms version 1.5.0 has been published on CRAN. Most importantly, it now offers the possibility to perform “custom summaries” (e.g. for a transformation of parameters or for a sum of parameters), and it now supports conditional-effects plots (see brms::conditional_effects()). See the NEWS for details. Of course, suggestions for improvements are very welcome!

4 Likes