Leave-one-out cross validation (LOO) for an astronomy problem

Harrison Siegel pointed us to this project with Maximiliano Isi and Will Farr on gravitational-wave analysis. The compare models using predictive evaluation, in particular leave-one-out cross-validation (LOO), as discussed here and here.

Siegel writes:

We discuss our implementation of the LOO in Sec. IIIC. Really hoping that we can motivate more people to use the LOO and move away from Bayes factors in my field.

Agreed. For further discussion of this point, see our article, Limitations of “Limitations of Bayesian leave-one-out cross-validation for model selection”.