František Bartoš

Why Companies Win by Investing in Open-Source Software - JASP Services BV In 2025, the IgNobel prize in Biology went to research on “whether cows painted with zebra-like striping can avoid being bitten by flies“. Suppose a fictitious company, ZebraCows, wishes to exploit th...

New blog post! Why companies win by investing in open-source software. With zebra-painted cows, of course.

www.jasp-services.com/why-companie...

We are pleased to have
@fbartos.bsky.social
join us today, Tuesday, September 30th, 11am (EST) to talk about Bayesian Hypothesis testing! This is followed by a workshop on using JASP for statistics around 12:10pm. The zoom is open to public with details in the flyer!
@PsychPrinceton

Simonsohn has now posted a blog response to our recent paper about the poor statistical properties of the P curve. @clintin.bsky.social and I are finishing up a less-technical paper that will serve as a response. But I wanted to address a meta-issue *around* this that may clarify some things. 1/x

> Why are you actively misrepresenting what others are saying all the time?

I'm happy to discuss with you in person if we meet anywhere, but I don't find replying to you online very productive at this point.

> Carter et al are right, and you are wrong

That's pretty much just arguing from authority

I did not say meta-analyses with huge heterogeneity lol. I said under any heterogeneity. Would you consider tau = 0.1-0.2 on Cohen's d scale with an average effect size of 0.2-0.4 huge? I would not. Pretty meaningful result (and probably representative of many meta-analyses), but p-curve fails.

> P-curve does what worse than random effects?

All the simulations I linked shows that p-curve estimates the effect size worse, on averate, than random effects.

Must've been a bug on the platform -- I could not see any responses I sent to the thread but other features worked fine.

For some reason, I cannot reply to Lakens anymore?

Regardless, if anyone is interested in the topic:
- Carter does not say something completely opposite to my claims
- I^2 is not a measure of absolute heterogeneity, Laken's argument strawmans meta-analysis
- p-curve does worse than random effects

It's not completely opposed - they say that they work well only under no heterogeneity. From their and other simulation studies it seems like that a simple random effects model performs better than p-curve even when publication bias is present. As such, I don't see any reason for using the method.

How is it directly opposite to what I'm saying?

Also, glad we got to the late-stage science when you start pulling arguments of authority. Always great debating with you :)

You are still free to find any third-party realistic simulations to address my claim :)

The issue is it fails even with low heterogeneity; you are just caricaturing any other slightly heterogeneous meta-analysis right now.

> And hey, even if a paper they wrote in 2014 on a new method is now partially outdated, so what?

I accept the critique and acknowledge the method is outdated and should not be used. It might have been a great idea back then but it did not turn out to be any more.

"Discovering Statistics Using JASP": The Promo Video - JASP - Free and User-Friendly Statistical Software The shiny new textbook “Discovering Statistics Using JASP” now has a 44-second promo video on YouTube, courtesy of Sage. What I like about the video is that it shows some of the pages, so you get an o...

Sage created a 44-second promo video for our book "Discovering Statistics Using JASP". Blog post: jasp-stats.org/2025/09/24/d...

The video itself: www.youtube.com/watch?v=MP0N...

Can anyone point me to the simulation studies showing that p-curve performs well under realistic conditions? And any done by someone else than pcurve authors? As far as I know, p-curve fails horrendously as long as any heterogeneity is involved...

doi.org/10.1177/1745...
doi.org/10.1002/jrsm...

btw, we just released JASP 0.95.2, which fixes some previously reported stability issues -- consider updating your version :)

The Bayesian part also provides more guidance on specifying prior distributions for estimation, testing, and model-averaging (with different effect sizes and in different settings).

Each manuscript walks you through three examples describing the applications of different meta-analytic tools, including

- effect size calculation
- funnel plot
- forest plot
- bubble plot
- simple meta-analysis
- meta-regression
- multilevel and multivariate models.

We released two preprints describing the JASP Meta-Analytic functionality in detail.

Meta-Analysis with JASP, Part I: Classical Approaches (arxiv.org/abs/2509.09845)

Meta-Analysis with JASP, Part II: Bayesian Approaches (arxiv.org/abs/2509.09850)

The methodology is implemented in the RoBMA R package
(update is coming to CRAN soon).

Vignette demonstrating the analyses:
fbartos.github.io/RoBMA/articl...

Preprint:
doi.org/10.48550/arX...

We derive posterior predictive distributions for many meta-analytic models. Importantly, meta-analytic models that ignore these discontinuities misfit the data and should not be used for inference; models that respect them provide a better basis for inference.

(see a couple of examples attached)

Publication bias is usually indicated by sharp discontinuities—typically at the significance threshold (selection for significance) or at zero (selection for positive results).

Similar plots are often used in metaresearch, we bring them to meta-analysis!

Z-curve plot is a new visual model fit diagnostic for #metaanalysis with an emphasis on #publicationbias. In contrast to funnel plots, z-curve plots
- visualize the distribution of z-statistics (where bias usually occurs)
- compare the fit of multiple models simultaneously

Publication bias is usually indicated by sharp discontinuities—typically at the significance threshold (selection for significance) or at zero (selection for positive results).

Similar plots are often used in metaresearch, we bring them to meta-analysis!

Figure 2 from the linked article.

Abstract of the linked article

New paper for anyone working with data:
Better data viz - for free, in few clicks.

Below, all N = 111, M = 0.04, SD = 0.27.
One-sided t-tests vs. 0 yield: t(110) = 1.67, p = .049.

Use raincloud plots. Or risk wrong conclusions!
Plot w/ @jaspstats.bsky.social today!

🧪 📊 #PsychSciSky #StatsSky

🧵👇

Fair Coins Tend to Land on the Same Side They Started: Evidence from 350,757 Flips Many people have flipped coins but few have stopped to ponder the statistical and physical intricacies of the process. We collected 350,757 coin flips to test the counterintuitive prediction from a...

See the published version of the manuscript for some coin-tossing fun-facts and interesting statistical modelling :)
doi.org/10.1080/0162...

Importantly, there are considerable differences in the degree of the same-side bias between our coauthor-participants. As the previous figure suggests, people can differ by 1-2%.

Further analyses showed that the bias seems to decrease over time, possibly due to practice effects.

We found that (as the title suggests) coins flipped by humans tend to land on the same side they started. The probability of the same side (50.8%) almost exactly matches the theoretical prediction of Diaconis, Holmes, and Montgomery from 2007.

Fair coins tend to land on the same side they started: evidence from 350,757 flips.

That's the title of our paper summarizing ~650 hours of coin-tossing experimentation just published in the Journal of the American Statistical Association.
doi.org/10.1080/0162...