As a result, reproducibility speaks more to the design and specification of the experiment than the truth value of a given result.

I'm sure @craiganthonyrs.bsky.social will be happy to adopt her.

Thanks, Andrea!

This kind of social media behavior genuinely bothers me. The thread is antagonistic in several places and misrepresents my position (assuming ignorance and confusion). It confuses my desire to speak broadly with a lack of domain knowledge.

Ouch, dude.

Gotcha!

I very sincerely want to protect the researcher here, so I don't want to give away what the paper is. Sorry if that makes it difficult to discuss.

I understand your point about small coefficients, though I would take this up as a separate issue. In the present case, these coefficients are effectively zero. They are each less than a percent of a percent of variance.

I never said anything about gross incompetence. Please don't put words in my mouth. My point is about standard inferential practices and how they fail us in edge cases. And how they failed an individual in this instance.

Thank you. All I can say is that my explicit motivation before writing word one was to criticize the process and try to protect the work (and the scientist). If I failed at that then that's my bad.

No, I didn't. I criticized the process that led to the work getting published. The work itself is not the point, hence my effort to protect the identity of the work.

Also, I wasn't criticizing anyone's work. I was illustrating a systemic failure in our field.

But the second thread misses the point of the first thread. The point is not to debate the mechanics of p values. The point is about reasoning about the work we do. Stats can't tell you whether your work is good or not.