10.1 In Example 10.1, apply a Weibull rather than Gompertz survival model to the mice death times. Compare the log pseudo marginal likelihoods of the two models and the chi-square statistic comparing actual deaths and posterior mean predicted deaths in the eight intervals of 100 days. The chi-square statistic under the Weibull should deteriorate, as early deaths are overpredicted and later deaths underpredicted.
10.2 In Example 10.2, adapt the BUGS code to include an intercept and using a car.normal prior for centred effects (constrained to sum to zero).
10.3 In Example 10.2, modify the piecewise exponential model to allow an interval specific treatment effect, . A random walk prior, analogous to that used for , may also be used as the prior on this varying treatment effect. Report evidence on the profile of the estimated and the changed LPML.
10.4 In Example 10.4 (marrow transplant survival), assess covariate effects and LPML resulting from a logistic assumption on the errors (log-logistic for ).
10.5 In Example 10.6 (first intercourse) compare the existing model with one ...