|
Abstract:
|
Maximum likelihood is a popular statistical method largely because it provides estimators with optimal statistical efficiency. However, many realistic statistical models are so complex in structure that it becomes computationally infeasible to find the MLE, especially in large data sets, such as DNA sequence data. One approach to solving this problem is the method of composite likelihood, which can reduce the complexity of computation at the price of some loss of efficiency. A composite likelihood is constructed by taking a product of likelihood terms, each one of which is a likelihood, conditional or marginal, for some subset of the data. In this talk, I will introduce the composite likelihood approach and the construction method from an optimal efficiency point of view. In addition, to illustrate the method I will consider its use in a recombination model for haplotype sequence data.
|
