A fit of the Poisson-gamma model to the pump-failure data quoted by Ge
orge et al. (Scand. J. Statist. 20, 147-156) using maximum conjugate l
ikelihood gives pump effects that agree closely with the values obtain
ed from a full Bayesian model. However, the model is a poor fit, becau
se the estimates of the pump effects do not look like a sample from a
gamma distribution. A split of the pumps into two subgroups does give
an adequate fit.