All,
We need a one-liner for Stan for inclusion in NumFocus materials. NumFocus is the non-profit that supports Stan. We could use a simple description else where too.

Deadline for resolution is Monday, July 3 but I suggest consensus by the previous Thursday, June 29 11 am EDT.

Some possibilities:

Stan provides portable tools for applied Bayesian statistical modeling, inference, and visualization.

Stan provides portable tools for applied Bayesian statistical modeling, inference, and visualization. Stan features a modeling language, math library, inference algorithms, evaluation, visualization, and interfaces in a variety of languages.

They want to use this for NumFOCUS collateral, which Gian Helfrich, describes as:

"NumFOCUS collateral" means the new website, a pdf/eBook we can distribute to potential partners & funders, a "media kit" to be sent to all project leaders to help with their own marketing & promotion, and general source content for describing NumFOCUS in a variety of channels.

That will work. Iâ€™d drop the â€śfrontiersâ€ť. And I wouldnâ€™t try to replace it with â€śstate of the artâ€ť or â€śsophisticatedâ€ť or anything else. I think â€śstate of the artâ€ť should be implied, and the other ones are either too boastful or too misleading that we only work on the frontiers. We want to be a go-to tool for day-to-day Bayesian analysis as well as pushing the frontiers.

Now if you want to write a Mission Statement, thatâ€™s where I think we want to be pushing the frontier and defining the state of the art.

My concern is that what we consider as â€śday-to-day Bayesian analysisâ€ť is the frontier in many if not most applied fields. Thereâ€™s also the issue of trying to differentiate us from other Bayesian software packages whose fragile algorithms donâ€™t allow anything but the simplest models. The huge advantage of Stan is that progression from simple to complex/sophisticated/state-of-the-art and in my opinion we want to emphasize that somehow.

I completely agree. But then I know what you mean by â€ścomplexâ€ť (bad adjective) and â€śsophisticatedâ€ť in terms of group-level effects, measurement error, missing data, complex likelihoods, informative priors, etc. But just writing â€śsophisticatedâ€ť sounds too vague and self-congratulatory to me.

Thatâ€™s just my opinion, of course, and Iâ€™m totally willing to compromise on the wording!

How about: â€śStan is a computing platform for Bayesian modeling, inference, and visualization that dramatically reduces the computational barriers that have limited the application of Bayesian methods in applied statistics.â€ť

Vote folks. Pick your top 2 in rank order. Poll closes noon EDT (New York Time) on Monday, July 3. -1 is a veto that explaining your thinking would be appreciated. Your veto is a vote.

Stan provides portable tools for applied Bayesian statistical modeling, inference, and visualization.

Stan provides portable tools for applied Bayesian statistical modeling, inference, and visualization. Stan features a modeling language, math library, inference algorithms, evaluation, visualization, and interfaces in a variety of languages.

Stan supports portable tools for applied Bayesian statistical modeling, inference, and visualization. Stan features a modeling language, math library, inference algorithms, evaluation, visualization, and interfaces in a variety of languages.

Stan is a platform for Bayesian modeling, inference, and visualization on the frontiers of applied statistics.

Stan is a platform for Bayesian modeling, inference, and visualization for applied statistics.

Stan is a computing platform for Bayesian modeling, inference, and visualization that dramatically reduces the computational barriers that have limited the application of Bayesian methods in applied statistics.

(2) and (3) are redundant. (5) is awkward in that the â€śappliedâ€ť should be moved up to go with â€śBayesianâ€ť.

(6) seems too historical, though I think it captures our mission.

I really do think we need something like a mission statement, which I completely believe to involve both removing barriers and pushing the frontiersâ€”theyâ€™re sort of the same thing, depending on whose perspective it is. Iâ€™m more excited that we make hard things possible than that we can bring Bayes to the masses. rstanarm, rethinking and brms are all doing a great job on the latter, though, and I donâ€™t think we should overlook that in the overall project mission.

We just have to make sure that weâ€™re then giving that community a path towards building their own bespoke models and really doing statistics as opposed to given them just another black box.

As a voting method I recommend score voting. Everyone gives a score on 0-10 to each one, and then add up all the scores, and take the one that gets the highest total. no vote = score 0

If no candidate is the first choice of more than half of the voters, then all votes cast for the candidate with the lowest number of first choices are redistributed to the remaining candidates based on who is ranked next on each ballot.[7] If this does not result in any candidate receiving a majority, further rounds of redistribution occur.

Suppose we should have a vote on what voting method and before you know it we have turtles all the way down.

Score voting is immune to Arrowâ€™s theorem (but not to general issues of strategic voting) because itâ€™s a cardinal (degree of approve) not ordinal (ranking) method.

I should not be trusted to pick voting systems. 3 way tie for first for first choice. Taking 2nd choices into consideration then 4) wins to my eye. Good exercise however and thanks for playing.