Friday, September 27, at 327 Yost
Refreshments: 3:00 - 3:30 p.m, Talk: 3:30
- 4:30 p.m.
It is now a widespread practice for researchers to
base conclusions about the inheritance of disease by
scanning large portions of the genome and conducting
a vast number of simultaneous hypothesis tests. A key
question to address is that of determining the probability
of a false rejection, that is, of incorrectly identifying a
genetic marker as linked with disease.
This talk will present a general
to approximating overall error rates using an importance sampling algorithm
given in Naiman and Priebe (2001). The basic idea is to generate
samples conditioned on the event that a hypothesis is rejected.
A standard genetic procedure known as the affected sib pair test,
will be described and used to illustrate the method.
Principles of efficient Monte Carlo simulation via fast Fourier transforms,
and concepts related to conditioning Gaussian samples, will be described.