How come my comment doesn't appear?

Harassing reviewers is actually not a normal part of science. In any case, I commented on your blog post.

I had plenty of criticism when I reviewed your paper, but then you started harassing me until Daniel told you to stop it.

At least not for you.

It's the exact same thing with you and Daniel: I try to have a discussion about statistical methods, but I just get accusations and insults.

As a scientist you should be concerned when your model conflicts with reality (in this case: there is no publication bias against significant clinical trials). You prefer to shoot the messenger.

EITHER you proved that there is a strong publication bias *against* significant clinical trials OR there is a problem with your method. The former is ridiculous, but you won't even consider the latter.

I'm pointing out a glaring problem with your z-curve method, and all you have to say is "do better"?

Despite n=41279, the estimated distribution of |z| between 0 and 1.96 is *completely* wrong. How reliable do you think the estimate is in a typical meta-analysis?

Here z-curve estimates that non-significant clinical trials are 1.66 times *more* likely to be published than significant ones. The "confidence interval" is 1.17 to 2.45. replicationindex.com/2020/12/24/i...

Hi Ben - I wrote that post. If you put your questions and comments at statmodeling, I'd be happy to respond there so that others can see!

Axioms don't have proofs.

Thanks for the nice blog post about our paper! I'll just add that our shrinkage estimator is a little bit ad hoc, but we did evaluate its performance: onlinelibrary.wiley.com/doi/full/10....

There's a nice discussion at Frank Harrell's datamethods:

discourse.datamethods.org/t/dichotomiz...

Hmm, I get 2x38 for 90% power of a two-sided test at level 10%. But nevermind. I came here to advertize an app which calculates the gain in power if you don't dichotomize. In this case, they would have needed only 2x19. The app is here: vanzwet.shinyapps.io/info_loss/