I'm a little disappointed that Bayes never gets mentioned here, however, since Bayesian methods are a systematic way to deal with all the problems in classical statistics that are talked about in this article (ie. stopping rules, "base rate")
Also there's no mention at all of the widespread abuse of parametric statstics -- that all the time people take some test out of a book that has 5 or 6 conditions that need to be true to use it and those conditions don't apply. There are many nonparametric techniques that are simple and foolproof and sacrifice a small amount of power.
It's 2012 and we can make short work of tough problems in a Bayesian framework with MCMC. There are good textbooks in Bayesian methods today, unlike the 1970s. It's time to send classical statistics packing.
I was tempted to introduce Bayes' theorem to explain the base rate fallacy; however, I couldn't devise a way to very quickly introduce it without just making it a magical formula I pulled out of my rear. It deserves more discussion than that. When I take a course on Bayesian statistics I may write a follow-up on alternatives to the flawed frequentist methods in the article.
Thanks for the tip about parametric statistics. I'll do some research.
2 comments
[ 6.0 ms ] story [ 13.3 ms ] threadI'm a little disappointed that Bayes never gets mentioned here, however, since Bayesian methods are a systematic way to deal with all the problems in classical statistics that are talked about in this article (ie. stopping rules, "base rate")
Also there's no mention at all of the widespread abuse of parametric statstics -- that all the time people take some test out of a book that has 5 or 6 conditions that need to be true to use it and those conditions don't apply. There are many nonparametric techniques that are simple and foolproof and sacrifice a small amount of power.
It's 2012 and we can make short work of tough problems in a Bayesian framework with MCMC. There are good textbooks in Bayesian methods today, unlike the 1970s. It's time to send classical statistics packing.
Thanks for the tip about parametric statistics. I'll do some research.