Chapter 8 Exercises
Dugongs

1. Take the BUGS code for modelling the growth of dugongs in Example 6.3.1, using uniform priors for alpha, beta gamma and log(sigma). Obtain posterior summaries and densities for each of these parameters, a plot of the model fit, and an estimate of the number of effective parameters via the DIC Tool. How does this compare with the true number?

2. Check the shape of the posterior distribution for gamma: do you think the uniform prior given to gamma is very influential? Try running again but giving gamma a prior more concentrated near one, say a Beta(8, 2). Does this make much difference? How might you summarise the effect of using a more 'informative' prior distribution?

3. Try fitting a growth curve that is quadratic in age. [Hint: it is helpful, for the sake of numerical stability, to standardise the x variable.] Does pD get the number of parameters approximately right? Is the model preferable to the original model according to DIC? Plot the fitted line to examine the fit visually. Is this a sensible model?