Nancy Reid
University of Toronto

Likelihood-based inference in complex models

Models for large or highly structured data often lead to likelihood
functions that are difficult to use for inference.  A variety of
likelihood-motivated approaches have been suggested, such as
quasi-likelihood, pseudo-likelihood, composite likelihood, pairwise
likelihood, etc.  I will survey some of the recent work on
likelihood-like functions in complex models, with a view to
understanding how these approaches are useful for inference, and what
their limitations are.