The Weight of Simplicity in Statistical Model Comparison

The epistemic weight of simplicity in science has, in the last fifteen years, been extensively discussed in the framework of statistical model comparison. This paper defends three theses: First, it juxtaposes Forster and Sober’s (1994, 2010) claims regarding the normative force of Akaike’s Theorem in statistical model analysis. Second, it elaborates the different targets of the various model comparison criteria, and shows that the weight of simplicity stands orthogonal to the issue of Bayesian vs. frequentist model comparison. Third, it proposes that statistical model comparison cannot be separated from the content of the original scientific problem, rendering elusive a neat division of labor between statisticians and scientists.