Comments on “probabilistic analysis of functional magnetic resonance imaging data”

Drs. Frank, Buxton, and Wong (1) present a useful Bayesian framework for statistical analysis of fMRI data. The Bayesian approach is promising because it enables one to include useful a priori information in the analysis. In this letter, I would like to point out a number of errors in the mathematics of their article. Most of these are minor and do not appear to effect the results and conclusions, while others are more important. Their Eq. [16] is incorrect. If we compare it with Eq. [4] and use the fact that