[R-meta] Some general reflections on the R-sig-meta-analysis mailing list

[Viechtbauer, Wolfgang (SP)]

Dear subscribers to the R-sig-meta-analysis mailing list,

It's been about 4 years since this mailing list was founded -- to be precise, on Tue 2017/06/06 15:05 CEST -- with the first post appearing on Jun 11 by yours truly (https://stat.ethz.ch/pipermail/r-sig-meta-analysis/2017-June/000000.html). Some of the reasons for starting this list are detailed in that post.

This list is managed by Guido Schwarzer, Michael Dewey, and me. This mostly involves dealing with posts from non-subscribers, since only people subscribed to the list can send posts to it -- for good reasons. Quite a bit of such non-subscriber posts are spam and we would not want this to go unfiltered to you. We also keep updating the filter to prevent repeated spam posts from the same address (although it's the usual whack-a-mole game). Some non-subscriber posts are also legitimate and we want those to go through, so we don't outright reject all non-subscriber posts. The amount of work involved with this is acceptable, at least so far. In any case, you are subscribed anyway, so thank you for that.

Speaking of subscribers -- at the moment, we have 438. I don't know the numbers for the other R mailing lists (https://www.r-project.org/mail.html), but leaving aside the 'big ones' (like R-help and R-devel), I would venture to guess that we are one of the larger ones, especially among the "Special Interest Groups". R-sig-mixed-models is probably larger and actually closely related -- since meta-analysis models are mixed-effects models -- but R-SIG-meta-analysis was founded to really put the focus on meta-analysis and the various packages and techniques for this purpose.

In terms of activity, I would say that the mailing list is also going well. The counts for the number of posts for each year so far are:

  year posts
1 2017   447
2 2018   907
3 2019   538
4 2020   670
5 2021   860

The number of people who post to the list is quite a bit smaller than the number of subscribers, but numerous individuals have expressed to me personally that they find the discussions of interest even if they do not actively participate and so they stay subscribed. Other people - unable to find the unsubscribe instructions - may have added r-sig-meta-analysis using r-project.org to their spam filter :)

The archives of the mailing list can be found here: https://stat.ethz.ch/pipermail/r-sig-meta-analysis/ It's a valuable resource as many discussions over the years have revolved around similar topics. Finding the right information/posts in the archives however is a challenge. Many search engines (like Google or DuckDuckGo) allow you to restrict your search to a particular site, so if you search for

site:https://stat.ethz.ch/pipermail/r-sig-meta-analysis

followed by your search term(s), you may be able to find relevant information more quickly. This will still require some digging though.

It would make for a nice (for some definition of 'nice') summer project to go through the archives and tag posts by topics or pick out threads that are focused on repeating themes (like dealing with dependencies under various circumstances, how to address publication bias in more complex models, and so on). I wish I had done this from the beginning on and now it would be a pretty time-intensive (but valuable) task.

At times, posts remain unanswered. This can happen for various reasons, starting with the most obvious one that people who could potentially answer just don't have the time at that moment. The question may also have been asked and addressed before, it may be unclear what the question is asking, and/or it may be a rather long post with many sub questions.

While I do spent a lot of time answering questions on the mailing list, these days I gravitate more towards questions that are relatively short, to the point, and focused on R itself. While the discussions on the mailing list often venture quite deeply into modeling theory, I actually think this should not be the primary focus of the mailing list. These questions are often way too complex to be discussed via a mailing list and instead should be discussed with a local statistical expert. I realize that such a person is often not available and places like the mailing list can be a rescue, but questions along these lines probably have a higher chance of remaining unanswered.

After searching the archives and making sure that the question or a similar one has not been discussed before, my general advice would be break up long questions into smaller posts, spread them out over some time, and try to abstract the problem away from your own specific case.

See also https://stat.ethz.ch/mailman/listinfo/r-sig-meta-analysis for some further instructions before posting.

Thanks for reading, subscribing, and participating,
Wolfgang