Skip to main navigation Skip to main content
The University of Southampton
Economic, Social and Political Sciences

Issues in construction of a composite likelihood, with a detailed biclustering example Seminar

Social Statistics and Demography
8 February 2013
Building 39 room 3013

Event details

S3RI Methodology Seminar Series

Composite likelihood has gained considerable momentum as a method of creating computationally tractable objective functions for a likelihood type analysis. We provide an example of a two way mixture model where the calculation difficulties make the problem intractable when there are many rows and columns. We then develop a composite likelihood that seems to capture much of the statistical information of the original problem. The composite likelihood is simple enough that we can repeatedly compute estimates to use for assessment of variability and for cross-validation for model assessment.

Speaker information

Bruce Lindsay , Pennsylvania State University. Eberly Chair in Statistics

Privacy Settings