You're viewing the discuss.systems public feed.

Wolfgang Viechtbauerwviechtb@scholar.social
May 6, 2026, 5:49 PM
@ianhussey Of course this can raise false positives (maybe in this example). To not annoy users, I would only issue this warning once per session. Also, the cutoff used can be adjusted via an argument. This still leaves the question what the default cutoff should be. There are already some papers around this idea: https://www.medrxiv.org/content/10.1101/2025.11.12.25340062v1 and tools based on this: https://tools.metapsy.org/reference/flageffectsizes Any comments on this general idea and the cutoff value to use? 2/2
💬 1🔄 0⭐ 0

Replies

Michael Deweymdewey@fediscience.org
May 7, 2026, 10:48 AM
@wviechtb @ianhussey my cynical self thinks that people will think that if their data passes the new test then everything is for the best in the best of all possible worlds and not actually look at their forest/funnel plot. In my experience reviewing for medical journals I see little sign of people looking at the plots anyway. You may have already seen https://onlinelibrary.wiley.com/doi/10.1002/ece3.73578 which shows the effort required to really check a meta-analysis.
💬 0🔄 0⭐ 0
Wolfgang Viechtbauerwviechtb@scholar.social
May 7, 2026, 11:57 AM
@mdewey @ianhussey A potential risk, but then I would say all hope is lost anyway (and it is unlikely that they would have actually read the docs to be aware of the fact that such a check was actually run).
💬 0🔄 0⭐ 0