Bob C-J and Geoff Cumming

New paper in press at JPSP! An adversarial collaboration focusing on a large-scale test of how strongly implicit racial attitudes predict discriminatory behavior. Pre-print here: osf.io/preprints/ps...

Smallest of interest is a good approach. But I do think, in general, that hypothesis testing should be for testing quantitative hypotheses, and that there is no shame in the careful observational and exploratory work needed to derive clear quantitative predictions.

Biggest predictor of success was sample size, but largely due to ability to detect trivial effects. We need to realize that a non-quantitative hypothesis isn’t developed enough for hypothesis testing— if you can’t yet specify an expected effect size, keep working until you can, and *then* test.

An unbiased tally of social science experiments shows less than 1/3 of hypotheses were supported. This is what our research literature *should* look like.

Biggest predictor of success was sample size, but largely due to ability to detect trivial effects.

Seems like some critical thinking about effect sizes is needed most of all: we shouldn’t still be pouring important resources into hypotheses that are just “more” or “different”. A researcher who doesn’t have a clear effect sizes expectation isn’t ready for a hypothesis test.

Conference Information Learn more about the conference!

Today's a good day to register for PsychTerms, an online workshop on the teaching of statistics and methods that will be held Dec 10-11: www.psychterms.com

Lots of great presentations to help pump up your teaching mojo for spring semester.

#stats

🔥

Download statistics for jamovi modules

Estimation is catching on!

The esci module in @jamovi.bsky.social was installed over 3,000 times in November. The R package in CRAN is getting an additional 250 downloads a month.

Are you estimation curious? So easy to get started with esci: rcalinjageman.github.io/esci/

#stats

We're not shocked at all the fake conference/paper inquiries we get but:

I got an email from an editor at Elsevier asking me to adapt a presentation they had seen me make at SFN.... a presentation I had *not* ended up making.

Why do we just shrug off routine mendacity like this?
#neuroskyence

And even for smaller studies! Especially to think critically about effects that yield p < .05 but with a ci that doesn’t rule out negligible effects… we need a better understanding that those call out for additional confirmation.

Also, please tag any of your other co-authors.

Tagging the lead author: @anthonystevend.bsky.social . Well done!

Here's a nice discussion of effect sizes in cognitive neuroscience, with real examples and an discussion of using null intervals. Lots of good advice here; if you've been interested in inference by interval, this is a great resource.

#Neuroskyence #stats

doi.org/10.1016/j.dc...

I couldn't easily find any of the authors in BlueSky -- happy to tag any authors, though? Great but deeply sad work.

And Nature Neuroscience ( @natneuro.nature.com )? You were an early adopter of reporting reforms! I believed in you. But anyone can see your authors fill in your reporting forms in their sleep, and that reviewers/editors ignore them. Such a shame. An editorial response in print? Seems needed.

I know finding statistical malpractice in the published literature is like shooting fish in a barrel, but this one still really makes me sad. The papers are from 2022... has the needle moved so little from the start of the reform efforts?

Quote from a Science article which made claims of correlation strength but only reported a p value: ("A strong correlation was also found with the AP upstroke/downstroke ratio of thepresynaptic cell only (P = 1.2 × 10–4 versus P > 0.3 for postsynaptic).)

The analysis examined correlation coefficients reported in 2022 issues of Science, Nature, and Nature Neuroscience.

Of 177 articles mentioning strength of correlation, 45% relied solely on point estimates; 53% only reported the p value to accompany these claims!

What are we even doing here?

The Prevalence of Misreporting and Misinterpreting Correlation Coefficients in Biomedical Literature Correlation coefficient is widely used in biomedical and biological literature, yet its frequent misuse and misinterpretation undermine the credibility and reproducibility of the scientific findings. ...

New analysis of statistical (mal)practice in top journals shows widespread problems in how correlations are reported and interpreted.

What is the word I keep reaching for that would mean: not surprised and yet still deeply shocked?

arxiv.org/abs/2511.14092

@nature.com,
@natneuro.nature.com

But adjusting his tie during a tank romp through a city… peak bond, in my opinion

Don’t know. My use is primarily reading for pleasure. I have Zotero on my boox and it seems to run well, but papers rarely win over my book collection so can’t really say.

Love the boox air. Sleek, powerful, great battery life. I use Calibre sync to pull my whole epub and pdf library from Dropbox onto my boox. It is awesome.

Another CREP study out in print: online.ucpress.edu/collabra/art...

Congrats to all the students and dedicated faculty mentors involved, like @aggieerin.bsky.social ky.social and @jorowags.bsky.social

Get your students involved in CREP: www.crep-psych.org!

#stats #psychology

thenewstatistics.com/itns/2025/09...

What We’ll Never Know
thenewstatistics.com/itns/2025/09...
Use these to lobby your politicians, not only if you are in the U.S. Suggest further makers of videos: tdw@virginia.edu

GitHub - dgbonett/vcmeta Contribute to dgbonett/vcmeta development by creating an account on GitHub.

New version of vcmeta (1.5) for R now out: github.com/dgbonett/vcm....

Some new functions and a few name changes.

vcmeta provides all the tools you need for variable-coefficient meta-analysis (no assumption of effect-size homogeneity; probably the best default option?).

#stats

Cool paper about JASP's raincloud plots with nice examples of how they can be used, especially for complex designs.

I do wish JASP would supplement add to this visualizations of group differences and their uncertainty.

#stats

doi-org.proxy.cc.uic.edu/10.3758/s134...

Donald Trump threatens to clear encampments in D.C. With D.C.’s MPD now under federal control and the National Guard coming to D.C., the fate of people in encampments remains up in the air.

"“I ain’t ever bought no prostitutes. I ain’t never raped nobody. I ain’t never paid anybody off. None of that stuff,” said G, a resident from one of the photographed tents.

“He’s much more of a criminal than I am”"

streetsensemedia.org/article/home...

Introducing JASP 0.95: Sandboxing, Plotbuilder, ESCI, Parametric Survival Analysis, and Much More - JASP - Free and User-Friendly Statistical Software We are proud to announce that JASP 0.95 has been released and is now available on our download page. JASP 0.95 presents a big step forwards, as you can see from the full release notes. Some of the hig...

Just out: JASP 0.95! With sandboxing, a new plot builder, ESCI, and more.

jasp-stats.org/2025/08/02/i...

Investigations of forgetting in Caenorhabditis elegans The traditional view considered forgetting as a passive process where memory traces gradually fade due to the natural weakening of neural connections.…

Great new review of forgetting mechanisms in C. elegans -- lots of cool research I had missed (or maybe forgotten about?)

#neuroskyence

www.sciencedirect.com/science/arti...?

Here's a nice paper in JCB on why testing for exactly no effect is not a great idea when studying biological systems, and how putting effect sizes at the front of our thinking can help.

Paper is by Josh Morgan at WashU St. Louis.

#stats #neuroskyence

rupress.org/jcb/article/...

Feel like an excuse to get that travel money for some sun and surf in November? Consider these two back-to-back meetings Down Under. 16-18 and 19-21 November 2025.
researchintegrityconf.com
aimos-inc.github.io/aimos.confer...