2 ms·
DD disappears when doing Bayesian model averaging: https://arxiv.org/pdf/2002.08791.pdf https://arxiv.org/pdf/2002.08791.pdf It seems DD is a phenomenon specifi
by yldedly 5y ago
DD disappears when doing Bayesian model averaging: https://arxiv.org/pdf/2002.08791.pdf https://arxiv.org/pdf/2002.08791.pdf
It seems DD is a phenomenon specific to point estimates.
>overfitting problem is really in terms of model flexibility versus data complexity, and not necessarily in terms of number of parameters per se versus sample size
Yep, well put.
- derbOac 5y agoThanks for the link! I hadn't thought about model averaging in this context but definitely bears on how to interpret this stuff. Interesting to think about too re: connections with model misspecification.