Biostatistics Seminar Series - Thomas Lumley, PhD

Title: "Pairwise likelihood for linear mixed models" Abstract: When fitting mixed models to multistage surveys there are two sets of structure to the data: the sampling structure defined by how the sampling was done and the model structure defined by shared random effects. When the structures are the same (or nested) it is often possible to ignore the distinction, but when they are potentially not nested the distinction is important. I am interested in inference for the model structure, with the sampling structure as a nuisance.  I will describe an approach based on pairwise likelihoods; an implementation for the nested case; and some simulations for the non-nested case.