Wednesday, November 15, 2006

Is God a Bayesian?

The moment I realized I have a job on hand, my blogging index has gone up tremendously!! Anyways,  this is a short conversation between me and Dave LeBaron.  Dave is a PhD student, one year my junior, at Penn State.

I guess the humor is slightly more math and meant for PhD's but to make it more readable, I will give you a brief background.  There are too dominant approaches to estimation in Statistics:  The classical approach that treats the data and parameters that govern how the data behave as different sets and develops point estimates for the parameters given the data.  The Bayesian approach in contrary treats both the data and the parameters as unknowns and estimates the distributions for the parameters.

Well, that's the best I could do.

Girish and Dave are walking down the corridor.

Girish:  Dave, if God were a Statistician, would he be Bayesian or Classical?

Dave: (Pauses for a moment) Well that's simple, he is surely Classical, because only the Devil could have come up with Bayes.

The icing to this story is, as you might know, that Thomas Bayes who first proposed the notion of conditional probability on which the entire Bayesian paradigm is built was an English priest.  The Devil turned out to be God's lieutenant !!!

No comments: