Brms::loo_compare() ; interpretation of WAIC deltas?

Interfaces brms
,
1

I am using the Oceanic Tool Complexity model described in Statistical Rethinking to evaluate my ecological data. I am deconstructing the model per https://bookdown.org/connect/#/apps/1850/access (section 10.2, search “bit by bit”) to calculate the WAIC for each variation of the model.

However, the 5 WAIC values are not very different from each other. How do I recognize if a sub-model’s WAIC value is meaningfully different compared to its related sub-model WAICs? Thank you.

ggsave.clusterAltModels.1.pdf (4.7 KB)

2

They are not meaningfully different, as all their uncertainty intervals are almost fully overlapping. If you saw some gap between the intervals for a model and those for another (such as between b10.12 and b10.14 in the WAIC plot of section 10.2) then you would have support for claiming a model to be better than another.