Operating System: Windows 10
Interface Version: rstanarm 2.15.3
I am comparing two models with 10 fold cv with each specified as:
stan_lm(form_b, prior = R2(0.5), data = models_rf) stan_glm(form_b, family = gaussian, prior_intercept = normal(0, 2, autoscale = FALSE), data = models_rf )
I get sensible coefficients for both models and no errors from stan. Diagnostics all look great, however when I run kfold with K = 10 I get
[] 10-fold cross-validation Estimate SE elpd_kfold NaN NA [] 10-fold cross-validation Estimate SE elpd_kfold -44.1 3.1
Any ideas what I’m missing? If K-fold = n then they ALL return NaN, and if I use loo I have the same issue regardless of k threshold.