3. In Example 9.4 (London borough suicides) apply the proper priors in (9.12) and (9.13)

to a model log(μi ) = β1 + si without predictors and compare their fit (e.g. by DIC). Also compare their consistency with the data by sampling newdata ynew and checking the extent to which the observed y are within 95% intervals of ynew. How do these proper priors compare in fit and consistency with the data with the full convolution model namely log(μi ) =

β1 + si + ui , with an improper CAR1 prior on s.