Conflicting inference from k-fold and loo-cv comparison

@vehtari I have mixed this post up with the many others I have running simultaneously. These last few replies (the ones below the reply I just deleted) belong to this thread dealing with ordinal vs gaussian regression for health outcomes, not linear vs spline for substance use outcomes. My apologies. I’m still interested in whether I got the syntax for the gaussian pit-ecdf calibration plot right though.