Supplementary Material for A re-evaluation of the model selection procedure in Pollet & Nettle (2009)

In this paper, we first explain the statistical model underlying the ordinal regression technique used by Pollet and Nettle (2009), including the two possible ways of calculat-ing the likelihood function (section 1). We then show that the model fit criteria reported were in fact invalid, and calculate the correct ones, showing that this leads to a different choice of best model (section 2). We then suggest two other strategies of model selection for these data, and show that these also lead to different best-fitting models than tha