#medstatgoe — Public Fediverse posts
Live and recent posts from across the Fediverse tagged #medstatgoe, aggregated by home.social.
-
📺 Video recordings of presentations from our recent workshop on "Recent Advances in Meta-Analysis" at @unigoettingen are now available at the University Medical Center's YouTube channel:
👉 https://www.youtube.com/@unimedizin_goettingenIncludes presentations by David Rindskopf, Annika Hoyer, Satoshi Hattori, David Jesse, Wolfgang Viechtbauer (@wviechtb), Renato Panaro (@panaro), Gerta Rücker, Bohua Chen and myself.
See also here for details:
https://medstat.umg.eu/aktuelles/symposium-meta-analysis-2026/ -
Meta-analyses may reveal effects not noticeable in individual studies, such as treatment-by-subgroup interactions (e.g. effects in males vs. females). To support Bayesian analysis, prior knowledge on interaction heterogeneity is crucial. @panaro compiled such data, showing heterogeneity is often lower than expected (which makes the derived prior valuable to support formal analyses).
See here for details:👉 https://arxiv.org/abs/2606.23968
(Joint work with @friede1)
-
🤔 Can you perform a meta-analysis of a single study?
💡 Yes, you can -- this makes perfect sense if you want to derive a "meta-analytic-predictive (MAP) prior" based on a previous study. And it is in fact done implicitly in certain shrinkage applications involving only 2 studies. @friede1 and I had a closer look at this case:
👉 https://doi.org/10.1017/rsm.2026.10081 -
📢 Our next #DFG -funded symposium on
"Recent advances in #MetaAnalysis"
will be taking place May 28/29 in Göttingen.
For more details, see here:
➡️ https://medstat.umg.eu/aktuelles/symposium-meta-analysis-2026/
and stay tuned for updates! -
🤔 Meta-analyses considering differences between subgroups within each study ("treatment-by-subgroup interactions") do not necessarily yield matching estimates for effects within subgroups and the difference between them.
💡 @panaro worked out how explicit consideration of information fractions contributed by subgroups in the analysis model allows to fix this counterintuitive behaviour; see here for details:
➡️ https://arxiv.org/abs/2512.18785
(joint work with @friede1). -
🤔 Estimation of between-study variability (heterogeneity) is tricky when only few studies are available.
❓ Can we make use of additional information, by considering subgroups within studies?
💡 It turns out we can -- yielding better performance due to fewer zero-estimates and more degrees-of-freedom.
👉 See here: https://arxiv.org/abs/2511.15366
(Joint work with Ao Huang and @friede1 ) -
📢 A mini-symposium on "Adaptive Designs - Modern Approaches in Clinical Research" will be held in Göttingen next February.
See here for more details:
👉 https://medstat.umg.eu/en/events/ -
Meta-analyses comparing treatment effects in patient subgroups (e.g. males vs. females) may be tricky, and may in fact sometimes suggest contradictory findings. Renato Panaro (@panaro) looked at the mechanisms at work, and also suggests a possible solution:
👉 https://arxiv.org/abs/2508.15531
(joint work with @friede1). -
Our Dortmund colleagues hosted an inspiring symposium last week within our #DFG -funded project on "Valid methods for meta-analyses with few studies and small sample sizes"; see here for details:
👉 https://msind.statistik.tu-dortmund.de/en/research/events/meta-analysis-symposium-2025
-
Most #MetaAnalysis approaches are based on parametric models, which may be challenging in case of sparse data (few / small studies). Thien Phuc Tran and Long-Hao Xu have developed a nonparametric permutation approach for #MetaAnalysis based on individual participant data (IPD); see here for details:
👉 https://arxiv.org/abs/2505.24774 -
🤔 Should one meta-analyze a single study?
☝️🤓 It turns out this makes sense in certain cases - if you're looking for predictions ("MAP priors"). In the particular case of a single study, there are also close links to power priors and bias allowance models. Have a look:
➡️ https://arxiv.org/abs/2505.15502
(joint work with @friede1 ) -
A very interesting workshop on "Hierarchical models in preclinical research" finished today in Göttingen. This was a joint undertaking of the IBS-DR working groups "Non-clinical statistics" and "Bayes Methods", and included an extensive Tutorial on #brms by Sebastian Weber and Lukas Widmer. Some of the material is available on the meeting website:
-
The EU-funded 🇪🇺 INVENTS project (https://invents-he.eu) held its first annual consortium meeting this week in Göttingen. Looking forward to further fruitful collaboration!
-
The working groups "Bayes Methods" and "Non-Clinical Statistics" of the International Biometric Society's German Region (IBS-DR) are organizing a workshop on
➡️ HIERARCHICAL MODELS IN PRECLINICAL RESEARCH
on December 5-6 in Göttingen.
Meanwhile the PROGRAM is available, and REGISTRATION is still possible -- see here for more details:
➡️ https://www.biometrische-gesellschaft.de/arbeitsgruppen/bayes-methodik/workshops/2024-goettingen.html -
A kick-off meeting of researchers from Japan and Germany within the #HeKKSaGOn network (www.HeKKSaGOn.net) came to a successful finish today in Göttingen. Looking forward to further collaborations!
-
Clinical trials are usually designed to demonstrate an overall treatment effect, but are often not able to resolve effects in patient subgroups. We performed a #MetaAnalysis, considering several studies to look for differences in treatment effects between patient subgroups in #MultipleSclerosis and investigate which patients may benefit most (or least) from immunotherapy. See here for more details:
➡️ https://doi.org/10.1177/20552173241274618 -
The working groups "Bayes Methods" and "Non-Clinical Statistics" of the International Biometric Society's German Region (IBS-DR) are organizing a workshop on
➡️ HIERARCHICAL MODELS IN PRECLINICAL RESEARCH
on December 5-6 in Göttingen.
See here for more details:
➡️ https://www.biometrische-gesellschaft.de/arbeitsgruppen/bayes-methodik/workshops/2024-goettingen.html