Here are my data and model:
m.brm <- brm(mean|se(sd) ~ 1 + (1+country||region) + (1+region||superregion) + superregion,data = datause, chains = 5, inits = "0", iter = 2000)
You can see the super region variable will restraint region variable.
However, I will get some impossible coefficient: like country Norway in Australasia：
Estimate Est.Error Q2.5 Q97.5
Asia Pacific, high income 0.38582809 1.465504830 -1.23782812 5.02839963
Australasia 1.03690937 3.100617094 -1.60713437 9.84860255
Europe, western 0.02773833 0.004218536 0.01809817 0.03522186
How should we adjust it?
Sorry it’s taken so long to reply. I don’t know much about what you are modeling, but a model like this does make a prediction about what would happen if Norwegian-Australian data were generated. The covariate combination is not possible in the real world, but the model doesn’t know that, nor is the model supposed to know that.
Tangentially related, your random intercept by superregion isn’t doing anything (except making the model hard to sample) given that you also have a fixed effect of superregion in the model. It would be a good idea to remove either the random intercept or the fixed effect.