University of Washington - Department of Statistics
The first speaker will examine a biological model where there are two types of data. In a sense, the relevant sample size for this problem is a two dimentional quantity. We will explore what the traditional idea of "asymtotic properties" for a model means in this context. Some examples will be given that indicate that if more data is collected, but of only one type, estimation may not improve. A condition on the relative rates of increase for the two types of data that is sufficient for consistency will be given.
The second speaker will describe some elementary recent examples where the classical MLEs and LRTs might be deemed unsatisfactory. These examples may have counterparts in model selection and latent variable problems.
Audience participation will be encouraged.