Following up on our recent post, I clicked on some of Ziliak’s links and found lots of good stuff, especially the post by Berk Ozler. I have no knowledge of his work but I like his writing; see here, for example.
Ozler’s post is very good indeed, and well written.
Ozler’s suggestion for improvement of the cowpeas trial does not include investigations of soil/weather/and other exogenous differences which will, in a random design, bias even the simplest difference, MS – TS, on a single farm. But this is to me a lingering trait of post-Fisherian agronomy, which has not looked into Gosset’s aka Student’s work on repeated balanced trials. Nor are current researchers aware as much as they might be of the battles between Student and Fisher, and the fact that Egon Pearson, Neyman, Harold Jeffreys, John Wishart and others sided with Student’s balance, not Fisher’s random.
Thus in my [Ziliak's] view, it is misleading to lump “Fisher, Yates, Neyman” et al together, and not mention Gosset. Lots of conflict between them. All were learning from and responding to Gosset’s work, and Fisher not in the most scientific manner (as I argue in detail in the Review of Behavioral Economics article).
I seem to recall someone (Rubin?) telling me that Gosset way overrated, but now I’m forgetting all the details.