Skip to content
- Colada [33] “The” Effect Size Does Not Exist (by Uri Simonsohn)
If I could choose one idea to persuade all researchers of, from all the ideas I have written about, it would be this one. So many papers, most meta-analysis papers for example, argue they are trying to estimate the evarage effect. Here i argue that’s a thing that does not exist
- Colada[80] Interaction Effects Need Interaction Controls (by Uri Simonsohn)
Explains a regression issue most people seem unaware of and demonstrates it with a paper written by Dahl & Card (the latter a Nobel prize winner), so clearly not something that’s widespreadly understood. If your key claim involves an interaction, in their case, that football losses cause more domestic violence when a team is expected to win, you need interaction controls, in their case, not just controlling for how many people watched the game, but that interacted with a loss.
- Colada[78] Drop that bayes (by Uri Simonsohn)
To propose Bayes factors are what researchers should be reporting, I believe requires not knowing what researchers are trying to do, or not knowing what Bayes factors actually do.
- Colada[120] Off-Label Smirnov (by Uri Simonsohn)
There is a well known classic statistical test, the Kolmogorov-Smirnov test. It has a test static “D”, which is not known to have an intuitive interpretation. But it does. It is the lower bound on the number of observations that differ across conditions.
- Colada[133.] Heterofriendly: The Intuition for Why You Always Need Robust Standard Errors (by Uri Simonsohn)
When I taught my first PhD-level methods course, I invited students to submit questions about any topic in statistics or methodology. Six out of 10 students asked about the same topic: robust & clustered standard errors. It’s clearly a topic they found both important and confusing.