Valencia: Summer of 1991

With the completion of the last edition of Jose Bernardo’s Valencia (Spain) conference on Bayesian statistics–I didn’t attend, but many of my friends were there–I thought I’d share my strongest memory of the Valencia conference that I attended in 1991. I contributed a poster and a discussion, both on the topic of inference from iterative simulation, but what I remember most vividly, and what bothered me the most, was how little interest there was in checking model fit. Not only had people mostly not checked the fit of their models to data, and not only did they seem uninterested in such checks, even worse was that many of these Bayesians felt that it was basically illegal to check model fit.

I don’t want to get too down on Bayesians for this. Lots of non-Bayesian statisticians go around not checking their models too. With Bayes, though, model checking seems particularly important because Bayesians rely on their models so strongly, not just as a way of getting point estimates but to get full probability distributions.

I remember feeling very frustrated and disillusioned at that 1991 conference, to see all these people who seemed to have no interest in going back to first principles and thinking about what they were doing. It’s like I tell our students: Grad school is the best and most open time. After that, most people are just stuck in their ways.

P.S. I’m not claiming any special virtue on my part. The above were just my reactions, and I’m sure that others since then have had similar reactions to my own mistakes.

3 thoughts on “Valencia: Summer of 1991

  1. Perhaps related, I searched WinBugs website this weekend for information about simulating priors by leaving out the data.

    All I found was a note from David S. around 2000 indicating numerical difficulties were likely to be encountered in WinBugs unless one used fairly informative priors.

    But almost all Bayesian analyses require multivariate priors and most analyses focus on only a few on these parameters (or a function of them). It’s not unusual for the priors for these parameters to be specified indirectly (i.e. implied by priors on other parameters) and many Bayesian analyses are undertaken using WinBugs so this should be a common undertaking…

    Perhaps if the posterior predictive checks look funny – but then those are probably seldom done anyways.

    K?
    p.s. guess I should buy the Robert and Cassella simulation book – but if others have suggestions or web material on how to simulate priors it would be convenient for me.

  2. … most people are just stuck in their ways.

    Bayesians always get together at the beach – Benidorm, Hamilton Island, Tenerife etc. Where better to stick your head in the sand?

  3. I am interested in exactly what you mean by 'first principles'. Is first principles coherence so that model checking is an application of coherence or instead is first principles some other general idea about science or statistics which suggests a weakness in the standard subjective Bayesian view (based on coherence)?

Comments are closed.