Michael G. Kenward

An application of maximum likelihood and generalized estimating equations to the analysis of ordinal data from a longitudinal study with cases missing at random.

Data are analysed from a longitudinal psychiatric study in which there are no dropouts that do not occur completely at random. A marginal proportional odds model is fitted that relates the response (severity of side effects) to various covariates. Two methods of estimation are used: generalized estimating equations (GEE) and maximum likelihood (ML). Both the complete set of data and the data from only those subjects completing the study are analysed. For the completers-only data, the GEE and ML analyses produce very similar results. These results differ considerably from those obtained from the analyses of the full data set. There are also marked differences between the results obtained from the GEE and ML analysis of the full data set. The occurrence of such differences is consistent with the presence of a non-completely-random dropout process and it can be concluded in this example that both the analyses of the completers only and the GEE analysis of the full data set produce misleading conclusions about the relationships between the response and covariates.

A taxonomy of mixing and outcome distributions based on conjugacy and bridging

Abstract The generalized linear mixed model (GLMM) is commonly used for the analysis of hierarchical non Gaussian data. It combines an exponential family model formulation with normally distributed random effects. A drawback is the difficulty of deriving convenient marginal mean functions with straightforward parametric interpretations. Several solutions have been proposed, including the marginalized multilevel model (directly formulating the marginal mean, together with a hierarchical association structure) and the bridging approach (choosing the random-effects distribution such that marginal and hierarchical mean functions share functional forms). Another approach, useful in both a Bayesian and a maximum-likelihood setting, is to choose a random-effects distribution that is conjugate to the outcome distribution. In this paper, we contrast the bridging and conjugate approaches. For binary outcomes, using characteristic functions and cumulant generating functions, it is shown that the bridge distribution is unique. Self-bridging is introduced as the situation in which the outcome and random-effects distributions are the same. It is shown that only the Gaussian and degenerate distributions have well-defined cumulant generating functions for which self-bridging holds.