Aaron Rendahl, University of Minnesota School of Statistics
11:30-12:20 on Wednesday, November 14, 2007
Olin Hall, Room 320
Lunch will be served
The likelihood ratio test is an important statistical method for comparing two competing hypotheses. While it is the theoretical justification for many standard statistical tests, it can also be useful for problems where standard tests are not applicable. This talk will focus on the mechanics of the likelihood ratio test and its application to the following experiment.
Unlike humans, haplodiploid species, such as the insects in the Hymenoptera order, do not have sex chromosomes. Sex is instead believed to be determined by whether or not the DNA patterns of the mother and father match at one or more sites. In some of these species, it is known that only one site need match, but despite begin first proposed in 1935, convincing experimental support for multiple site matching has been lacking. This research inbred eight generations of the parasitoid wasp C. vestalis, and showed that the model where two sites needed to match fit the data far better than the model where only one site needed to match. Comparison of the models was performed using the likelihood ratio test, and a p-value was obtained using Monte Carlo simulation.
This research was conducted by Jetske deBoer and colleagues at the Department of Entomology at the University of Minnesota, and analyzed by the author while working in the Statistical Consulting Clinic of the University of Minnesota.