Epic self-citation (the ONLY citation) but this looks like a major cautionary exposition for fields where expected effect sizes may be very small (eg genetics).
Thanks for pointing this out. It's a very interesting-sounding article, will have to read, but I don't know if my stats are strong enough to judge the article. Is SSRN a high-impact journal?
physics : ArXiv :: social science : SSRN
i see ;)
What do you expect from Taleb?
He is amazingly arrogant. Possibly several nanoDijsktras
Given that the statistician's mantra is that no model is true, it's just that some are useful, I would expect to see a sensitivity analysis of any claim being made as to the effect of choosing the wrong model on the accuracy of parameter estimates or predictions. Instead, what we get here is mathematical smoke and mirrors, 'cause ultimately what the paper says is that even when you have a very large sample, people might reasonably disagree about how to model that sample and that the data itself won't be sufficient to decide which model is best -- hence "statistical undecidability". Okay, yeah, sure, interesting point, but not exactly shocking to anyone who has ever used any sort of statistics.