I've copied my tweet history over to 🦋BlueSky. Please find me there!

@JasonRubinstein

RT @sharoz: Reminder that I rarely check twitter anymore. BlueSky is my primary social media site. BlueSky: Blog:…

Reminder that I rarely check twitter anymore. BlueSky is my primary social media site. BlueSky: Blog: LinkedIn:

@GidMK Interesting. That puts the p-value almost exactly where it's expected.

@GidMK Gotcha. The value-vs-error distinction is what I didn't understand. But can I suggest that it's clarified in the abstract and paper? The abstract seems to suggest that all small p-values are errors.

@GidMK Something is definitely wrong. I'm not arguing about that. But I don't think it's accurate to describe a p-value below 0.001 as an "error". The problem is the absurd effect size. But any study with a moderate effect size that oversamples is going to have a sub-threshold p-value.

@GidMK Yes, but having the context of how uncommon it is for that ES+N is important. Just as 2e-35 is uncommon, so is p>0.001 for that ES+N. Would you consider a p-value of 0.01 an error too? Why consider one tail an error but not the other?

@jamesheathers I should add, I agree that some p-value can be improbably small. But the confidence interval or distribution of expected p-values needs to be computed for a given sample size and effect size.

@jamesheathers I don't know the field well enough to say. But if the argument is that effect sizes in the field are small, it's important to include that context in the preprint. So maybe add "for such small effect sizes and sample sizes" in prominent places like the abstract.

@jamesheathers Redoing the graphs. Extremely small p-value are likely under even moderate effect sizes and reasonable sample sizes.

@RaphaelWimmer I try to minimize responding to twitter people for any purpose other than convincing to leave. Let them feel the isolation for staying here. On mastodon, I'm at:

RT @sharoz: BlueSky June! Many people who have a BlueSky account still prioritize posting, reading, and asking questions on Twitter. So f…

@ChelseaParlett @cshperspectives @dggoldst @dsquintana @EikoFried @jamesheathers @NicoleCRust @RuthRosenholtz @sTeamTraen @stephenjwild @tcarpenter216 @AlexKale17 Great! Just asking people to give another shot now that it's more feature complete. Make sure to check out the StatsSky feed and use #Stats for anything relevant.