Dr Mircea Zloteanu 🌺🌞🍃Apr 13

#524 Interactions in Logit Regressions: Why Positive May Mean Negative

Thoughts: Nonlinear models require a bit more caution for new researchers when making inferences.

#logit #inference #causalinference #interaction
#datacolada

https://datacolada.org/57

[57] Interactions in Logit Regressions: Why Positive May Mean Negative - Data Colada

Of all economics papers published this century, the 10th most cited appeared in Economics Letters , a journal with an impact factor of 0.5.  It makes an inconvenient and counterintuitive point: the sign of the estimate (b̂) of an interaction in a logit/probit regression, need not correspond to the sign of its effect on the...

Data Colada
Bjørn SætrevikJan 13
I appreciated this episode of @TimHarford's #CautionaryTales this morning, which compared the #PiltdownMan hoax to the #DataColada vs. Gino/Ariely debacle: https://timharford.com/2025/11/cautionary-tales-homo-deceptus-sciences-dirty-little-secret-2/
Dr Mircea Zloteanu 🌺🌞🍃Jun 19, 2025

#statstab #369 Don't Trust Internal Meta-Analysis

Thoughts: I don't agree with all the assumptions, but interesting to know how bad these can be.

#metaanalysis #QRPs #minimeta #datacolada #falsepositive #replication

https://datacolada.org/73

[73] Don't Trust Internal Meta-Analysis - Data Colada

Researchers have increasingly been using internal meta-analysis to summarize the evidence from multiple studies within the same paper. Much of the time, this involves computing the average effect size across the studies, and assessing whether that effect size is significantly different from zero. At first glance, internal meta-analysis seems like a wonderful idea. It increases...

Data Colada
Show threadRich Stein (he/him)Dec 27, 2024

@davidho.bsky.social
This statement is loaded:
"Similar to the rise of Wikipedia before, it seems many people appreciate the access and convenience of ChatGPT, even if they know it cannot be trusted entirely."

Imagine a world where the error bars simply no longer matter — science journals where replication is not an issue because" good enough "is good enough.

We've been thinking about this the wrong way the whole time. 🙃
#DataColada

Matt HodgkinsonNov 14, 2024

The Francesca Gino case ended with her being found guilty of data manipulation, exposed as a serial plagiarist, and losing her defamation suit against the Data Colada bloggers.

Here's a very watchable 15-minute summary by Pete Judo: https://m.youtube.com/watch?si=DW5gVlr364Nzb4ta&v=Q9tgyVPytBk

#ResearchIntegrity #FrancescaGino #ScientificFraud #HarvardUniversity #Plagiarism #DataColada

It's Over - Gino vs Harvard Fake Data Scandal

YouTube
Rich Stein (he/him)Sep 13, 2024
"Court clears researchers of defamation for identifying manipulated data"
https://arstechnica.com/science/2024/09/court-clears-researchers-of-defamation-for-identifying-manipulated-data/ v @arstechnica
h/t @0xabad1dea
https://infosec.exchange/@0xabad1dea/113128835175584035
#DataColada #FrancescaGino #HBS #Harvard
Court clears researchers of defamation for identifying manipulated data

Harvard, however, will still face trial over how it managed the investigation.

Ars Technica
Mikka, MDSep 13, 2024

@datacolada is free :)

https://arstechnica.com/science/2024/09/court-clears-researchers-of-defamation-for-identifying-manipulated-data/

I fear a world in which science isn't based on conduct and validity but courts anymore. I look forward to a world in which Harvard gets smacked hard for its conduct in this case. #datacolada

Court clears researchers of defamation for identifying manipulated data

Harvard, however, will still face trial over how it managed the investigation.

Ars Technica
Ben WaberMay 18, 2024
For those like myself who missed this, the good folks at #DataColada have an update on the legal wackiness with Gino's meritless lawsuit https://datacolada.org/116
[116] Our (First?) Day In Court - Data Colada

“Any update on the lawsuit?” That is the most common question any of us is asked. It is usually preceded by an apologetic preamble, like, “sorry if this is a sensitive question,” or “I don’t know if you’re tired of talking about this, but…” The reality is that, for the most part, our actual sensitivity...

Data Colada
Felix SchönbrodtMay 10, 2024

Update from #DataColada on the $25M defamation lawsuit #Gino vs. Harvard vs. the DataColada guys.
https://datacolada.org/116

They report on the 1st step of the trial: The case could be dismissed if there's no basis for a defamation. Defamation requires that it is not an opinion but a (false) fact; furthermore, it makes a difference whether Gino is considered to be a public figure, and whether DataColada „knew the statement was false“ or they were reckless ...

Read the details in their blog post!

[116] Our (First?) Day In Court - Data Colada

“Any update on the lawsuit?” That is the most common question any of us is asked. It is usually preceded by an apologetic preamble, like, “sorry if this is a sensitive question,” or “I don’t know if you’re tired of talking about this, but…” The reality is that, for the most part, our actual sensitivity...

Data Colada
Show threadBen WaberFeb 28, 2024
Next was a fantastic talk by Don Moore on the prevalence of overconfidence and overprecision at #DataColada. Moore shows these psychological tendencies in a variety of settings, and also describes how different visualization approaches can reduce these biases. Highly recommend https://www.youtube.com/watch?v=jvscL_btKIo (6/8) #psychology
Don Moore (Univ. of California, Berkeley) - Data Colada Seminar Series (01 May 2020)

YouTube