16.8 Summary

First model with groups!

• observations on one group are independent to another group but correlated in the same group

• New parameters : group-specific

• global parameters

• learning for one group to another will lead to some shrinkage

• this models are less variable than no pooling and less biased than complete pooling