Servan L. Grüninger

F**k my life

Goose chasing meme. "What's your research question?" Then angry goose chasing yelling the same phrase.

Stats consulting is constantly like:

Them: I want you to run (complex stats)

Me: OK, what's your research question though?

Them: ...A (complex stat)?

Me: Research question?

Them: You know, like the stats in this journal article. Something reviewers will like.

Hannah Arendt Prize: "Comparison is the way we know the world" Read the talk that Masha Gessen gave at the ceremony of the Hannah Arendt Prize in Bremen.

Masha Gessen:

“And this is why we compare. To prevent what we know can happen from happening. To make "Never Again" a political project rather than a magic spell. And if we compare compellingly and bravely, then, in the best case scenario, the comparison is proven wrong.”

Maybe that‘s a feature, not a bug. I have the feeling that having a public audience does not always bring out the best in us…

#scisky: What are (at leat somewhat) reputable scientific endeavours / research programmes of today that scientists of tomorrow might deem utterly unscientific?

Interesting, thanks! Why do you think so (for any of them)?

Analysis | OpenAI and X: Promises of populist technology, shaped by a single man In the end, power in Silicon Valley is largely held by a small cadre of believers and loyalists who deliberate in secrecy and answer to no one.

These technologies "are supposed to be so democratized and universal, but they’re so heavily influenced by one person... Everything they do is [framed as] a step toward...larger greatness and the transformation of society". But they're just "cults of personality”. Well said, Noah Giansiracusa

#scisky: What are (at leat somewhat) reputable scientific endeavours / research programmes of today that scientists of tomorrow might deem utterly unscientific?

When conditioning on prior activity on the data is indeed needed to make a test valid, overlooking that a procedure should be modified to accommodate this prior activity might lead to an erroneous test. However, this situation only arises if we disregard the elementary principles of statistical inference such as correct conditioning, sufficiency, completeness and ancillarity. Conditional inferences are statistically valid when their interpretation is properly conditioned on the information extracted from the observed data, which are sufficient for model parameters. Therefore, unconditionally stating that double-dipping, data peeking, or using data more than once invalidates inference does not make statistical sense. In contrast with common reform narratives, one can use the data many times in a valid statistical procedure. Below, we describe the conditions under which this validity is satisfied. We also discuss why preregistration cannot be a prerequisite for valid statistical ...

Box 2. Valid inference using data multiple times. We assume a test based on an unbiased test statistic generates valid inference, in the sense of achieving its nominal Type I error probability, under its assumptions within the Neyman–Pearson hypothesis testing paradigm. Information extracted from the data prior to the test of interest is represented by a statistic from prior analysis. Cells describe the necessity and/or the outcome of conditioning the test of interest on this statistic from prior analysis, for varying levels of information captured. Some technical clarifications for special cases are discussed in appendix C. Left. The statistic from prior analysis is not used in decision making, for example, by combining it with a user-defined criterion which might affect aspects of the test of interest. Many commonly used linear models fall in the first column where procedures are based on an optimal test statistic and therefore, using the information from prior analysis does not aff

Saw some revived discourse on twitter about using data twice (why not more?!!) and came here instead to remind you that we talk about it in this paper.

Don't believe when someone says "you can't do that"! Ask "why?"

🧪 #metasci #stats
royalsocietypublishing.org/doi/10.1098/...

We reviewed 100 psych. simulation studies & find room for improvement in planning/reporting. As a remedy, we (František Bartoš, @timpmorris.bsky.social Anne-Laure Boulesteix, @danielheck.bsky.social & Samuel Pawel) present ADEMP-PreReg, a simulation study preregistration & reporting template 🧵/1

1/ Statistics do not simply reflect reality, but reconstruct it.

A recent episode in Switzerland illustrates this nicely. The Swiss Federal Statistical Office made a mistake: They incorrectly aggregated the results of the most recent parliamentary elections.

Falsche Wahlergebnisse: Das Debakel um falsch zusammengezählte Wählerstimmen bei den eidgenössischen Wahlen offenbart eine tiefer liegende Wahrheit: Statistiken bilden die Wirklichkeit nicht einfach ab, sondern konstruier...

3/ The episode serves as a cautionary tale about the power of statistics and the responsibility of those who use them. #stats #chvote

More (in German): www.nzz.ch/feuilleton/b...

2/ For three days, politicians and pundits discussed the political consequences of a statistically created reality that eventually turned out to be false.

1/ Statistics do not simply reflect reality, but reconstruct it.

A recent episode in Switzerland illustrates this nicely. The Swiss Federal Statistical Office made a mistake: They incorrectly aggregated the results of the most recent parliamentary elections.

Retire the P-Curve (On Twitter) Collective behavior from fish to fascists.

Ooooh I almost forgot this fantastic blog post by @jbakcoleman.bsky.social
joebakcoleman.com/blog/2023/pc...

I can recommend the resources provided by Penn State: online.stat.psu.edu/statprogram/...

New paper with @ianhussey.bsky.social @taymalsalti.bsky.social @rubenarslan.bsky.social. Most measures are only used once or twice, w/o agreement on gold standards - a serious barrier to cumulative science. that leads to lots of papers without meaningful progress. www.nature.com/articles/s44...

Imo, pre-registration can help in two ways in this context: 1) It helps researchers involved in a particular analysis to think about theory *before* getting started. 2) It makes it easier for other researchers to spot issues with the analysis.

Theory matters. Models matter. Assumptions and assumption violations matter. Methods are not equivalent in their assumptions or statistical guarantees. We do not have the luxury of making inference in a vacuum. That's also why reproducibility/replicability doesn't mean what people think it means.

I misread it as "How to make a Friend on BlueSky" and got really hopeful for a split second

#HiSciSky!
Here for #stats and #metascience, especially in #biology and #biomedicine.

Somebody stop him, he’s gone mad with power!