Comparison of two models in one script and see which one provides best fit

Interesting. I’ve long wondered why there seems to be preference among the experts for separate-models-and-loo-comparison over the mixture approach. How is it that the former skirts the closed-world assumptions of the latter?