search

paul-buerkner / brms

brms R package for Bayesian generalized multivariate non-linear multilevel models using Stan
https://paul-buerkner.github.io/brms/
GNU General Public License v2.0
1.28k stars 184 forks source link

Stan folks have invited you to write a blog post #99

Closed wfolta closed 7 years ago

wfolta commented 8 years ago

Paul, this has appeared on Andrew Gelman's blog, under an article on Stan:

I added brms and added asterisks to indicate that brms and rethinking are not packages that we (the Stan dev team) wrote or are maintaining. If whoever wrote brms (or someone else) wants to write a blog post and mail to the stan-users list about what it does, you can send it to me (carp@alias-i.com) and I can post it here on Andrew’s blog.

paul-buerkner commented 8 years ago

Thanks Wayne! I have talked to Bob and I will prepare a blog post in the next few weeks.

jgabry commented 8 years ago

Indeed. The link to Wayne's blog post on brms and rstanarm was put on the blog a while back but I think you've added a bunch to brms since then so maybe time for an update! No rush, but anytime you have something written up that could work as a blog post let me know and I can put it in the blog queue (but we can skip Andrew's standard 2-3 month delay). Sounds like Bob's happy to do it too, so feel free to let either or both of us know.

On Wednesday, July 13, 2016, Wayne Folta notifications@github.com wrote:

Paul, this has appeared on Andrew Gelman's blog, under an article on Stan http://andrewgelman.com/2016/07/12/29511/:

I added brms and added asterisks to indicate that brms and rethinking are not packages that we (the Stan dev team) wrote or are maintaining. If whoever wrote brms (or someone else) wants to write a blog post and mail to the stan-users list about what it does, you can send it to me ( carp@alias-i.com javascript:_e(%7B%7D,'cvml','carp@alias-i.com');) and I can post it here on Andrew’s blog.

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/paul-buerkner/brms/issues/99, or mute the thread https://github.com/notifications/unsubscribe/AHb4Q42EWLBFO9BTn5DKTTdI0_AWChozks5qVYASgaJpZM4JL-Ol .

paul-buerkner commented 8 years ago

I want to wait until brms 0.11 is released because then I can also talk about distributional regression (i.e. also predicting family specific parameters such as the residual standard deviation 'sigma'). Additionally, I can mention the Bayesplot package in this case so that it recieves some initial awareness.

wfolta commented 8 years ago

Ooo, you keep on impressing me. I realize rstanarm has the mission of acting a lot like the default functions that people already use, but brms is so incredibly flexible. I hadn't fully though through the implications of the mixed ordinal logistic regression that I've been doing lately, and how specialized that really is -- but simple with brm.

jgabry commented 8 years ago

Yeah our goal with rstanarm really is about providing robust defaults, with moderate flexibility, abd a particular focus on guiding people through the transition to the Bayesian workflow.

What Paul has done is to provide a legitimate alternative to manually programming in Stan a wide (and growing) array of fairly complicated models, along with convenient and familiar methods for working with the results. It's such a cool package and really an awesome addition to the Stan universe.

Jonah

On Wednesday, July 13, 2016, Wayne Folta notifications@github.com wrote:

Ooo, you keep on impressing me. I realize rstanarm has the mission of acting a lot like the default functions that people already use, but brms is so incredibly flexible. I hadn't fully though through the implications of the mixed ordinal logistic regression that I've been doing lately, and how specialized that really is -- but simple with brm.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/paul-buerkner/brms/issues/99#issuecomment-232534013, or mute the thread https://github.com/notifications/unsubscribe/AHb4QyV7OqhbUUi7vZh29h0Qa4PzYy6Iks5qVZMhgaJpZM4JL-Ol .

wfolta commented 8 years ago

Jonah, Yes, rstanarm provides an excellent bridge. As another example, it starts sampling immediately -- pre-compiled models -- which saves the 30-45-second compile that brm has. For someone just starting, extra compile time (particularly for a simple model) could be off-putting, so rstanarm does well there. (And its complex models do some impressive tricks as well to maximize performance.) The flip-side is that brm can do most anything, and you can take brm's code and build on it or learn from it.

Both excellent projects, thanks to all involved! (I just want to make sure brms gets some press as well.)

jgabry commented 8 years ago

On Thursday, July 14, 2016, Wayne Folta notifications@github.com wrote:

Both excellent projects, thanks to all involved! (I just want to make sure brms gets some press as well.)

Absolutely. The more attention we can bring to projects like brms that make Stan and the perspective on statistics we promote accessible to a wider audience the better!

And personally, it's a great feeling to see the Stan community expanding in this way. That is, to know that we have people like Paul and Richard (rethinking) -- and everyone else who is developing Stan-related tools or materials for teaching Stan (including your blog posts Wayne!) -- who value what we're doing and want to help bring it to as many people as possible.

jgabry commented 8 years ago

Also, Paul I plan on recommending brms when Andrew and I teach our intensive Bayes and Stan short course next week here in NYC. Anything in particular you want me to mention that I might not know about (eg new features coming soon)?

On Thursday, July 14, 2016, Jonah Sol Gabry jgabry@gmail.com wrote:

On Thursday, July 14, 2016, Wayne Folta <notifications@github.com javascript:_e(%7B%7D,'cvml','notifications@github.com');> wrote:

Both excellent projects, thanks to all involved! (I just want to make sure brms gets some press as well.)

Absolutely. The more attention we can bring to projects like brms that make Stan and the perspective on statistics we promote accessible to a wider audience the better!

And personally, it's a great feeling to see the Stan community expanding in this way. That is, to know that we have people like Paul and Richard (rethinking) -- and everyone else who is developing Stan-related tools or materials for teaching Stan (including your blog posts Wayne!) -- who value what we're doing and want to help bring it to as many people as possible.

jgabry commented 8 years ago

Feel free to email me separately about that as I guess this is a github issue I'm commenting on and this isn't an "issue".

On Thursday, July 14, 2016, Jonah Sol Gabry jgabry@gmail.com wrote:

Also, Paul I plan on recommending brms when Andrew and I teach our intensive Bayes and Stan short course next week here in NYC. Anything in particular you want me to mention that I might not know about (eg new features coming soon)?

On Thursday, July 14, 2016, Jonah Sol Gabry <jgabry@gmail.com javascript:_e(%7B%7D,'cvml','jgabry@gmail.com');> wrote:

On Thursday, July 14, 2016, Wayne Folta notifications@github.com wrote:

Both excellent projects, thanks to all involved! (I just want to make sure brms gets some press as well.)

Absolutely. The more attention we can bring to projects like brms that make Stan and the perspective on statistics we promote accessible to a wider audience the better!

And personally, it's a great feeling to see the Stan community expanding in this way. That is, to know that we have people like Paul and Richard (rethinking) -- and everyone else who is developing Stan-related tools or materials for teaching Stan (including your blog posts Wayne!) -- who value what we're doing and want to help bring it to as many people as possible.

paul-buerkner commented 8 years ago

Thanks for mentioning brms in you short course! I think that distributional regression will be the main new feature this month apart from incorporating the bayesplot package.

jgabry commented 8 years ago

I didn't even end up doing a single rstanarm example (we just entirely focused on Bayesian concepts and writing Stan programs) but I put brms and rethinking with rstanarm on the list I gave people with recommendations for what packages and materials to check out after the course.

On Thursday, July 14, 2016, Paul-Christian Bürkner notifications@github.com wrote:

Thanks for mentioning brms in you short course! I think that distributional regression will be the main new feature this month apart from incorporating the bayesplot package.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/paul-buerkner/brms/issues/99#issuecomment-232721969, or mute the thread https://github.com/notifications/unsubscribe/AHb4Q-Y9rmCtAEmonZbP3sYvBbabJro9ks5qVmb7gaJpZM4JL-Ol .

paul-buerkner commented 8 years ago

Thank you @jgabry!

paul-buerkner commented 7 years ago

The blog post is now available at http://andrewgelman.com/2017/01/10/r-packages-interfacing-stan-brms/

Thanks @jgabry for making this possible!

jgabry commented 7 years ago

No problem. Glad we go the post up. Keep up the great work!

On Wed, Jan 11, 2017 at 3:59 AM, Paul-Christian Bürkner < notifications@github.com> wrote:

The blog post is now available at http://andrewgelman.com/2017/ 01/10/r-packages-interfacing-stan-brms/

Thanks @jgabry https://github.com/jgabry for making this possible!

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/paul-buerkner/brms/issues/99#issuecomment-271813966, or mute the thread https://github.com/notifications/unsubscribe-auth/AHb4Q_SEfn4qGsH7lodXywROP4bMNEmuks5rRJoPgaJpZM4JL-Ol .