xkcd: Frequentists vs. Bayesians

I just KNEW it would end up in MGo.Lico.us

http://andrewgelman.com/2012/11/16808/

"All statisticians use prior information in their statistical analysis. Non-Bayesians express their prior information not through a probability distribution on parameters but rather through their choice of methods. I think this non-Bayesian attitude is too restrictive, but in this case a small amount of reflection would reveal the inappropriateness of this procedure for this example."

David Berri is still the dude on the left.

I don't know what historical vindictive I'm wading into here but, honestly, this attack on Prof. Berri has me rather perplexed.

I'm a very proud, bleeding-blue Michigan alumnus, an avowed Bayesian and also a contributor to Dave Berri's Wages of Wins blog.  In fact, I'm one of two really strong Bayesians that contribute to the site (the other has his own site titled "Bayesian of Wins").

While I wouldn't classify Dave as a Bayesian himself, he is very far from the religiously-blind Frequentist that Munroe depicts in his cartoon.  Dave has welcomed our attempts to add some bayesian priors to his analysis (check out wagesofwins.com for some recent examples) and he never falls back on the absolutes of empirical analysis.  His approach is more like "This is what the data seem to tell us. However, we should be careful not to over-extrapolate."

While it is most definitely essential not to look at empirical output without context, as Gelman points out above, you also can't ignore the empirical data.  You can't even avoid it.   You just have to make sure to put it in context.  The Dave Berri that I know never shies away from this.