03/29/2016 - 12:15pm

03/29/2016 - 1:10pm

Speaker:

Mark Huber (CMC)

Abstract:

In statistics, a contingency table is a vector of values $x$ that is subject to linear inequalities $Ax \leq b$. Usually the vectors are integer valued. In order to conduct hypothesis testing for tabular data, the Monte Carlo approach requires the ability to sample uniformly (or at least approximately uniformly) from these contingency tables. In this talk I'll discuss a Markov chain approach to obtain such samples, and some simple ways to deal with nontrivial constraints.

Where:

Millikan 2099, Pomona College