Date: Feb 21, 2013 8:01 PM
Subject: Re: Trying to understand Bayes and Hypothesis
Maybe you guys have an idea on how to help me then. I have been working on a paper and it is very unpopular in my field. Maybe as bitter as the Frequentist/Bayesian split but it is purely a statistics issue.
Truthfully, I have always thought the problem wasn't with Frequentism per se, but on the scientist side. I have read the polemical arguments both sides have written and, but I have tended to fault the actual practice of science and not the tools.
I am making a contentious argument. In my field we have a set of theories which are generally considered normative theories. The only problem is that they don't work empirically and to make matters worse, they handed out two Nobels for them. My argument is that there is a math error deep inside the theories, that the likelihood function must be the Cauchy distribution for the Bayesian form and that the test statistic must be the Cauchy distribution on the Frequentist side. Because my data set is so large, the Bayes factors are so outrageous between accepted theory and my model that I think they are being rejected by readers as they are not p<.01. The numbers are orders of magnitude away from what is found as rejection values on the Frequentist side.
Theory says the "errors" should be normally distributed and no one argues that a variety of goodness of fit measures reject it at p<.001 or wherever the table stops.
Theory says I should be able to minimize variance choosing an expectation or maximize an expectation choosing a variance. Of course you cannot do that with a Cauchy distribution.
Since I am finishing my grad school studies any POLITICAL suggestions on this? I was even cussed out during a conference presentation. As the person was a polyglot I actually learned quite a bit.
Likewise any suggestions on how to get it published. I am being desk rejected out the ears.