Stan General Meeting Thursday, December 10, 2020, 11 am EST

Hangouts Link: https://meet.google.com/gzm-wmum-pfm

Instructions: Ask to attend in the hangouts interface and someone should let you in in the first 10 minutes of the meeting. Email Breck if you have problems or want to attend the physical meeting in New York City when they start again.

Meeting Agenda:

  1. Introduction to the Liverpool group of Bayesian modelers–will cover during 1 min intros.
  2. Update on alternate hosts for general meeting.

Please add your agenda items in replies.

1 Like

Agenda item: Supporting Stanc2 in next release

With some recent changes in the Math library we can no longer guarantee models most will compile with the old stanc2. They are actually more likely to not compile in C++ then they are. Which begs the question if supporting the “backfall” option is still useful?

shouldn’t this be discussed in the dev forum?

this comes up at every release - the answer seems to always be that it depends on where rstan is at.

1 Like

Yes, but I just wanted fast feedback on that and if it would be “no” on the meeting we dont need to debate further and waste time.

I know, but this time I am asking because this week we merged a Math/Stanc3 change that severly reduces how many models would compile with stanc2. models with a user defined function for which any of the arguments are vector/row vector/matrix would most likely not compile now. So mostly only non-complex models work now, for which I am not sure we need the backfall.

I am not saying we must remove now but at least add a warning that it may no longer work should be added.

1 Like

Oh… well… then it sounds as if we can leave stanc2 behind us. The thought with rstan is probably anyway not useful? I mean, waiting for rstan to be able to cope with the stanc3 parser would only make sense if the stan services are useful for a mixed rstan release which relies on the stanc2 back port and a new stan services - as would be the case if we have huge improvements to the sampler. As this is not the case we can probably just drop stanc2 support.

1 Like

Topic for discussion: Would it make sense to add some mechanism by which it is easier to discern who has contributed to each example in the Stan User Guide, enabling folks in the forums to tag those contributors most likely to be able to help when looking to understand or extend the examples?

1 Like

you can always look in GitHub

I’ve tried that, and it’s not a scalable solution. Possibly more context would clarify: often there are posts in the forum where a user is trying to understand or extend one of the examples in the SUG. When I see such posts and don’t know the answer myself, it would be convenient to know who among the devs and other forum volunteers might have pertinent expertise so I can tag them in. We have an existing private spreadsheet maintained by @martinmodrak listing folks’ general domains of expertise that is intended for this purpose (as well as keep track of who is currently overloaded and therefore load-balance; though I’m not sure if this feature still works), but the terms folks have put there are rather broad. There seemed to be consensus in the meeting that we should keep using this private spreadsheet (something public may induce a deluge of DMs) and simply add a column where folks can enter the SUG sections they’d be comfortable advising on.

if you want to let new people play the game, you need to give them room to speak up. calling out the “experts” every time something comes up deprives folks of the opportunity to become one.

Few things to add to this:

Thanks for bringing this up

It just requires me to do a minor cleanup about once a month. When I don’t forget, it works smoothly.

This is an interesting idea and if multiple people are up for this, I think we can give it a go.

I’ll however say that my experience with tagging people is mixed at best: I think the actual rate of tagged people engaging is definitely less tahn 50%.

This is an important point. I think this is mitigated by using this tagging only for the questions that didn’t get an answer organically. I generally (when I am less overwhelmed than the past ~two months) try to monitor the forums for users that are newly active and responding and try to create some point of contact. If we keep using the spreadsheet I may also get back to asking them to be added to it. Once again the experience is mixed as quite a few people agreed to be put on the list (and I tried to ask with as little pressure as possible) but many (most?) of them never really reacted when I tagged them.

(aside: if you notice a new user that became at least a bit regularly active, let me know!)

I still didn’t figure out a good way to monitor this (if I knew someone really consistently is not responding to this it is better for everybody to remove them, but I don’t know how often others tag etc…)

So these are my thoughts now… Hope they make sense.

1 Like