11. In Example 11.10 (second model) adopt a reduced model with autocorrelated ei jt excluded, but with multivariate normal and multivariate t (via scale mixing with unknown degrees of freedom) priors for the clinic effects (b j1, bj2, bj3). Do these models improve on the fit of the independent prior model, and are any unusual clinic effects detected by the scale mixture approach? Finally consider the model yi jt = bj1 + bj2t + bj3Bi j + ηN + ηA + wi j + ui jt , where bj2 have means mj2 that are modelled in terms of patient treatment (so differential gain by treatment can be assessed).