Goodness of Fit from Algebraic Invariants

Speaker:

Jason Morton– Department of Mathematics, Pennsylvania State University

Description:

Many statistical models for discrete data have a set of algebraic invariants that vanish on any probability distribution which is a member of the model.  We discuss a new idea for performing goodness-of-fit tests on these models using such algebraic invariants.  This approach may work better near singularities and boundaries, and in other situations where the usual approaches such as the likelihood ratio test have problems.

Event Topic

Nonlinear Algebra and Statistics (NLASTATS)