The Journal of Data Science has a couple of interesting Bayesian papers in the April 2006 issue.
The first article addresses a thorny topic, multiple comparisons in an ANOVA model.
- A Bayesian Approach to the Multiple Comparisons Problem. Andrew A. Neath, Joseph E. Cavanaugh. Journal of Data Science 2006: 4(2); 131-146. [Abstract] [PDF] (Model, Bayesian)
The approach that Neath and Cavanaugh take is quite intriguing. Multiple comparisons represent an effort to discern whether certain pairs of means are different from one another or whether they are the same. You can formulate this as a series of Bayesian models and exam the posterior probabilities for these models. The authors also examine Bayesian model averaging to produce estimates of individual means for each group.
The second article discusses the teaching of Bayesian statistics.
- Training Students and Researchers in Bayesian Methods. Bruno Lecoutre. Journal of Data Science 2006: 4(2); 207-232. [Abstract] [PDF] (Model, Bayesian)
This article is written from the perspective of someone who believes that anyone who thinks carefully about statistics can't help but abandon traditional frequentist methods because of their many flaws. The language is quite blunt. Users of frequentist methods
must make "judgmental adjustments" or "adaptative distorsions" designed to make an ill-suited tool fit their true needs.
Dr. Lecoutre advocates the use of non-informative priors and offers the term "Fiducial Bayesian" to describe this approach. Although I dislike the aggressive tone of this article, it is worth reading.