st: Statalist in a forum format

[Stas Kolenikov <skolenik@gmail.com>]

Dear Stataliasters,

thanks again to Marcello Pagano, Stata Corp side team and everybody
involved in setting up statalist.org. I however would like play a
devil's advocate and argue that we DO NOT NEED statalist in that form.
I would instead argue that we need to find ways to move the statalist
discussion to the StackExchange system (http://www.stackexchange.com).

Let me describe the system in bold strokes. (Disclaimer: I am closely
familiar only with the statistics website on StackExchange family,
http://stats.stackexchange.com, although I keep an eye on another five
or so sites, such as latex, math, biking, and board games, out of 100+
for all walks of life). On each of the sites, users ask questions and
seek inputs from other participants. Other users can answer these
questions. (So far, that's trivial, and every forum does that. The
crucial differences begin at this point.) The quality of different
answers can be compared by upvotes or downvotes that registered users
can cast. Answers can be improved by editing. In fact, questions can
be edited, as well (so at the very least, one could edit a novice's
question to incorporate the code formatting and tag the topic(s) that
the question raise(s)... and NJC can edit Stas Kolenikov's questions
for silly grammar :) ). Besides full answers, users can also provide
smaller-scale comments. Duplicate questions are being identified,
flagged, and closed, with links to what they are duplicating. A user's
contribution to the site is measured by their "reputation" that is
accumulated through good questions and answers, and by badges (asking
questions roles, answering question roles, editing roles, flagging
roles, etc.). Thus there is a variety of feedback mechanisms that help
improving the quality of questions and answers, and motivate the users
to participate. Thematic sites provide special formatting and/or
syntax highlighting, like Latex-type MathJAX formatting on math or
statistics sites (implemented on statalist.org, as well); code
formatting on programming and statistics sites (partially implemented
on statalist.org, requires some cleanup, as far as I understand; on
stats.stackexchange.com, R code is provided with simple syntax
highlighting, too); chess board rendering on the chess site; card
rendering on the poker website; etc. There's a sophisticated system of
tags that have mini-wikies (available on mouse-over) and full wikies
(available on clicks); users accumulate their reputation both overall
and within tags. Moderators play a major role in keeping the site
running, as it is the moderators who take actions on low quality
questions and answers (close duplicates, migrate questions to more
appropriate sites, etc.)

The biggest difference between a typical forum format (including
statalist) and StackExchange is the conceptual vision of the mission
they have in mind. Historically, plain-email-text statalist has been
aimed at "User 1 posts their current problem" - "User 2 posts a
solution to that specific problem" - "User 3 refines, if there's room"
- and everybody forgets about the whole thing a month later (although
in the case of statalist, we often see referrals to statalist
archives). StackExchange works on generating knowledge about the
universe (in this case, this would be Stata universe) that is supposed
to stay for prosperity, so that website users and the Internet at
large could benefit from the previously generated answers. The
ultimate goal is that StackExchange posts show up on the first page of
Google when you search something like "fixed effects model with
heteroskedastic panels". In some of my Google queries, I do indeed get
SE posts, so somehow the mechanism of StackExchange I described above
does work.

I am not quite sure how to better describe my next point, but I think
that creating statalist.org compartmentalizes Stata as an obscure
software. While those in the academic ivory towers can enjoy their
academic freedom, others in industry and government have to fight this
impression at their workplaces ("How is it called again? Why are you
not happy with some alternative software that we have a license
for?"). This is a tiring uphill battle. (Literally today, I had to
convert some SAS code to Stata and back, and StackOverflow, the
programming branch of StackExchange, offered some good pointers...
which I got from the first Google search page!) The statistics branch
of StackExchange, CrossValidated (http://stats.stackexchange.com), is
heavily entrenched in R, and the popular claims regarding popularity
of R are based on the counts of posts on StackExchange
(http://r4stats.com/articles/popularity/). Moving to stackexchange.com
rather than to statalist.org would help strengthening the argument
that Stata is a proper, kosher, mainstream statistical package. A
fivefold or tenfold increase in the volume of Stata questions that
could happen once hsphsun2 closes would make a good splash in Stata
reputation in the big world outside statalist walls.

Toning this down a little bit, I think the best of two worlds can be
combined on statalist.org if the major components of the StackExchange
system are utilized. The unique features that Stata Corp can donate to
the system to make it superior over the limitations of the format and
capabilities of StackExchange would be syntax highlighting similar to
that of the Do-Editor, and automatic links to the online help files
(http://www.stata.com/help.cgi?whatever) that are added once a Stata
command is being identified in the code section. A more prominent use
of tags is a must for the forum to be usable as a reference, rather
than a hit-and-run (ask-and-wait) system that it has historically
been. Stata commands may act as tags (based on either what user enters
as a special tag if that's something user-contributed, like -outreg-
or -ivreg2-, or what is being recognized automatically, which I
suggested above). Users should be required to use tags, or at least
asked to confirm when they are submitting a question with no tags. (7
tags on 78 "General" forum questions, and two more from "Using the
forum"??? I'd say it is not working yet.) For tags to work well,
though, other users should be able to edit tags in the original posts.
Adding any of that functionality would require serious work on the
forum engine. How much the engine can be expanded, I have no idea; I
am not a website programmer. If this equates to just reproducing
StackExchange engine, it is obviously a futile exercise.

Joining StackExchange system as a separate site may or may not work
out (see http://area51.stackexchange.com/faq -- there should be a
critical mass of users, questions and topics, and so far, I am aware
of about five or so statalist people who are active on CrossValidated,
which won't create a critical mass). R questions are split between the
statistics website and the programming website, depending on whether
it is more of "why doesn't lmer produce a standard error for variance
components" (statistics) or "how do I tweak ggplot2 to produce
transparent graphs" (programming). I imagine Stata questions could
work pretty much the same way, too. At the moment, there is about 1
question a day on Stata on statistics website, and may be two
questions a day on programming website... compared to 10+ new
questions on statalist... on a slow day.


-- Stas Kolenikov, PhD, PStat (ASA, SSC)
-- Principal Survey Scientist, Abt SRBI
-- Opinions stated in this email are mine only, and do not reflect the
position of my employer
-- http://stas.kolenikov.name
*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/