Supplement 1. OpenBUGS code for statistical models.

File List openbugscode.txt Description The file openbugscode.txt contains the code for the models used for re-analysis of Hawkins et al. MTE data – and was used to analyze the 23 linear data sets (by the authors' criteria) only. The response variable lns[] is log species richness, and the explanatory variable mtetemp [] is temperature transformed according to the MTE formulation explained in Hawkins et al. The variable group [] indexes the data set. The parameterization of the multivariate normal priors on the regression parameters (alpha [,]) follows suggestions in Peter Congdon (2003), Applied Bayesian Modelling, Wiley and Sons: Chichester. To obtain the results reported in the paper, the model was run for a burn-in period of 10,000 samples followed by a run of 10,000. Three chains were run, starting with different initial values, and the samples were thinned by 10 to reduce autocorrelation. Convergence was checked using the BGR disgnostic in OpenBUGS (v. 2.2.0, Feb. 2006).