If the appeal doesn’t work, try publishing elsewhere. If your criticism is strong, it will find a home. The original journal will look as if they are more interested in their reputation than in getting the science right. I’m doing a similar thing now, though not regarding a meta-analysis.
A graph showing that mean probability judgments that a streak of similar outcomes would continue were below the base rate for most terminal streak lengths greater than one, consistent with the gambler's fallacy
Revised preprint!
Science may be self-correcting, but only if the corrections are published.
A 2025 Psych Science article reported no gambler's fallacy for probability judgments w/ truly random sequences, but Yeonho Choi and I found strong evidence for the fallacy in the authors’ data.
Honored to have my in-press paper featured by SJDM!
Thanks for the plug! I just submitted the (first round of) page proofs.
New preprint!
A recent Psych Science article reported no gambler's fallacy for probability judgments w/ truly random sequences and a disconnect between such judgments and binary predictions.
Yeonho Choi and I reanalyzed the data and found very strong evidence for an effect and a much smaller gap.
I will need to change one of my examples for the German cities task.
It is interesting that two of the data points in the small data set were from that part of Texas.
A graph of predicted loss of life as a function of warning time from flash flood when there are 1,000 people at risk. With no warning, over 130 fatalities are expected. With one hour, only about 10 are expected. Those numbers are for "high force" floods with swiftly moving water. Note: the data are more than 30 years old.
It's a shame the county decided not to install a warning system (e.g., sirens) along the Guadalupe River. Here's a 30-yr-old graph showing the benefit of warning time for forceful floods w/ pop. at risk = 1000. Would love an update based on more data.
Paper: onlinelibrary.wiley.com/doi/abs/10.1...
Nice paper w/ some great points. Maybe see Michel Regenwetter's papers on generalizing to "people." Other examples: Hershey & Shoemaker (1980, reflection effect) and - self-serving - one of mine (2016, do multiple plays eliminate certainty/possibility effects?). Still a role for btw-Ss exps, imho.
A table listing some limitations of meta-analyses and how metastidies address them. Metastudies provide advantages related to applicability; time, effort, and money; publication bias; statistical precision and power; generalizability; and the interpretation of results for moderators.
Wonderful paper! For more on purposive variation, see this 2022 paper on metastudies: journals.sagepub.com/doi/10.1177/...
Free postprint: osf.io/preprints/ps...
It's like Simonsohn et al.'s new mix-and-match method, but with more attention to moderation, generalizability, and statistical power.
When completely described, the certain option includes a negation, as in "200 people will be saved and 400 people will not be saved."
Some theories have no place for such negations and so don't make clear predictions. Others predict no framing effect in choices between completely described options.
New PsyArXiv preprint! osf.io/preprints/ps...
Risky-choice framing effects persist when option descriptions are matched in gains and losses and even when the options are completely described.
(Last year's Psych Science results replicated nicely in a larger census-matched sample!)
Teaching tidbit for JDM/risk perception: NPR story on a fire in a battery storage facility is a perfect example of Slovic's "signal potential." Link in comment.
“Ultimately, the incident has tremendous potential to derail the industry, not just within California, but across all of North America.”
I was there! It was raining the whole time, and people were undeterred.
An appeal to anyone conducting a "megastudy": Treat it like a big counterbalanced experiment (as in a "metastudy"; see link to a not-so-new paper), so you don't end up with some of the same problems as meta-analyses, such as having the features of the interventions being correlated and imbalanced.
An appeal to anyone overseeing a "megastudy": Treat the whole thing like a big counterbalanced experiment (as in a "metastudy"; see the link), so you don't end up with some of the same problems as meta-analyses, such as having the features of the interventions being correlated and imbalanced.
I'm surprised the X logo has not spouted little clockwise feet to resemble a swastika. Graphic designers, where are you?
Wonderful paper! At the risk of appearing/being self-serving, here are three papers on perceptions of energy and water use (not perceptions of others' beliefs) that used similar methods (mixed models):
www.pnas.org/doi/full/10....
www.sciencedirect.com/science/arti...
www.pnas.org/doi/full/10....
Our daughter McKenzie (not pictured) is one of the fired Presidential Management Fellows featured in this story. Feel free to share.
DOGE has never been and never will be about efficiency. There is zero chance that firing these talented, dedicated young employees improves government efficiency.
Living liver donors as well, though not the topic of the film.
Trump/Musk fired our daughter McKenzie. A smart, dedicated servant who has worked for the US Census, the US Embassy in Bolivia, and the US Treasury. Fulbright Scholar, Presidential Management Fellow, and (until today) a probationary IRS economist. She moved to Utah for the job in November. Fuckers.
Is anyone tallying the number of federal employees fired in each state? It seems potentially useful (if anything is) to be able to tell a Senator that Trump has fired, say, 500 of their constituents. Or that Trump has fired, say, 200 Ohio State graduates.
What source article are you talking about?
I don't know about love, but the movie dialog provides a great example of the sunk-cost effect, useful for teaching (or testing) JDM. Jason Alexander's character says something like, "Back out? Forget it, pal! We've got a thousand man-hours in on this."
BTW, we're running another one right now, using four scenarios where we found a significant effect in recent studies (in a particular condition) and four scenarios from earlier studies in which the authors found a nonsignificant effect in the same condition. The data are trickling in.
You might like the metastudies approach, with stimuli and other aspects of the design varied to learn about generalizability and boundary conditions: journals.sagepub.com/doi/abs/10.1...
We also have follow-ups for framing effects with extreme probabilities, somewhat risky options. See PsyArXiv.
