Results of the 4 models being fit to the data.

Lower DIC is better, and higher Bayes factor is better, so Cue Mixing has the best score in both metrics. Bayes factors are ± 0.96% (95% CI).
aDIC is perhaps not properly defined for a model with no variable parameter space. If the choice probability is considered a parameter, then it has no variance, and thus

D
-
 = D(
θ
-
)

. Reported here is simply the deviance, or equivalently, the DIC with pD = 0. The values of pD were 0, 9.49, 1.13 and 1.81, respectively.
bThese are not independent metrics. These numbers are presented to show that the priors have been adjusted to make the posterior roughly equivalent, since we need a large number of samples for each model in order to make an accurate estimate of each Bayes factor. The percentage error in estimated Bayes factor for each of the 4 models is expected to be very similar, within 0.15%, when we have 100,000 samples and these posterior weights.
Results of the 4 models being fit to the data.