tag:blogger.com,1999:blog-10448119.post4927227363060189840..comments2024-03-23T06:42:53.608-04:00Comments on Hogg's Research: unjustified uncertainty estimatesHogghttp://www.blogger.com/profile/18398397408280534592noreply@blogger.comBlogger1125tag:blogger.com,1999:blog-10448119.post-66800792341195091872010-05-21T16:24:49.468-04:002010-05-21T16:24:49.468-04:00Most uncertainty estimates in the literature are u...Most uncertainty estimates in the literature are unjustified. It's common practice to assume some model for the prior probabilities (priors AND sampling distributions) that is a pathetic representation of anyone's prior beliefs. E.g. Gaussian noise.<br /><br />I think this comes down to the misconception that probability can deal with "random" errors but not "systematic" errors. There aren't really two kinds of errors. Errors are just "the amount by which your data are off relative to the thing you were trying to measure". There's no point doing an analysis and quoting a "formal statistical error" when you KNOW that p(D|theta) was a crap model of your prior beliefs.Brendon J. Brewerhttp://www.physics.ucsb.edu/~brewer/noreply@blogger.com