|
Abstract:
|
Correlated data, such as multivariate or clustered longitudinal data, arise commonly in practice. It is often of interest to understand the marginal response process and the degree of association among responses. Often, those data contain missing values in response vector, and this causes complexity for valid inference. We propose a robust approach for incomplete clustered longitudinal data using composite likelihood. Specifically, using pairwise likelihood we describe how to carry out robust estimation with minimal model assumptions involving the mean and association structures. We also show that the resulting estimates remain valid for a wide variety of missing data problems including missing at random data and so in such cases there is no need to model the missing data process.
|
