Trek separation in Gaussian graphical modelsAlgebra & Discrete Mathematics
|Speaker:||Kelli Talaska, UC Berkeley|
|Start time:||Fri, Nov 12 2010, 2:10PM|
This talk will explore a connection between combinatorics and algebraic statistics. In particular, we will look at Gaussian graphical models, whose covariance matrices can be given in terms of certain path families called treks. Inspired by classical results in algebraic combinatorics, we develop a graph-theoretic criterion for determining the rank of a submatrix of the covariance matrix. (No prior knowledge in statistics will be assumed.) This is based on joint work with Seth Sullivant and Jan Draisma.