Not exactly sure why we're using the distribution of (1/lamda) here rather than just plain lamda in part (iv) - as far as I can see it doesn't explicitly state for you to use this, happy to be corrected!
Thanks John for your help with these and the others - I'd much rather ask stupid questions on here than sit in the exam hall wishing I'd asked them! I think I get this now, the E[lamda] would just produce a Bayesian estimator of lamda. So in general, we need to look at the expectation through the posterior of how the mean is defined in the original density function (1/lamda here) to get an estimate of the mean of the density function taking into account prior beliefs?
