Invitation from Steve Goodman (sgoodman@jhmi.edu)

1 Nov 2004

As you know, I am statistical editor at Annals of Internal Medicine. They are very keen on getting a good, accessible review/tutorial on the issue of validating clinical prediction rules. In particular, they would like insight into the issue of internal validation re-using the whole sample (e.g. Bootstrapping, jacknife), versus spit-sample techniques. Old stuff, but the clinical research community really doesn't understand the this, and apparently there is some influential formal guideline from who-knows-where that explicitly states that split-sample techniques are superior to resampling. Anyway, they asked me who could write such a piece, and you immediately came to mind. Any interest? This would likely get high visibility in the medical research community because of the journal's profile. You would be free to write this w/a colleague of your choosing, or if you wanted to collaborate w/me, I'd be happy to help as well (although I could not take the lead). Possible? Feel free to call me with any questions.

Thanks!

Regards, Steve

---

5 Nov 2004

Frank - Looks like a cosmic convergence here. I checked out the page (what is a "wiki" page anyway?) and it looks terrific. I don't know if I have anything to add, but I will look more closely at it. I think my biggest contribution may be organizational, although you and David are both pretty experienced at writing for audiences like this. I note that you are trying to cover a lot of territory here, but we would probably accept a decently long piece if that what it took to cover the material properly. I would put an upper limit somewhere near 5K words, optimally somewhat less. I would also say that the relative emphasis in this journal should be at least equally on clinical prediction rules versus array data since Annals sees very little of the latter, even though genetics is certainly where much of the action is these days.

Steve

---

Steven Goodman, MD, MHS, PhD
Associate Professor of Oncology, Pediatrics, Epidemiology and Biostatistics
Department of Oncology, Division of Biostatistics
Johns Hopkins Schools of Medicine and Public Health