Abstract
We study exponential families of distributions that are multivariate totally positive of order 2 (), show that these are convex exponential families and derive conditions for existence of the MLE. Quadratic exponential familes of distributions contain attractive Gaussian graphical models and ferromagnetic Ising models as special examples. We show that these are defined by intersecting the space of canonical parameters with a polyhedral cone whose faces correspond to conditional independence relations. Hence serves as an implicit regularizer for quadratic exponential families and leads to sparsity in the estimated graphical model. We prove that the maximum likelihood estimator (MLE) in an binary exponential family exists if and only if both of the sign patterns and are represented in the sample for every pair of variables; in particular, this implies that the MLE may exist with observations, in stark contrast to unrestricted binary exponential families where observations are required. Finally, we provide a novel and globally convergent algorithm for computing the MLE for Ising models similar to iterative proportional scaling and apply it to the analysis of data from two psychological disorders.
Funding Statement
This research was supported through the program “Research in Pairs” by the Mathematisches Forschungsinstitut Oberwolfach in 2018. Caroline Uhler was partially supported by NSF (DMS-1651995), ONR (N00014-17-1-2147 and N00014-18-1-2765), IBM and a Simons Investigator Award.
Piotr Zwiernik was supported by the Spanish Ministry of Economy and Competitiveness (MTM2015-67304-P), Beatriu de Pinós Fellowship (2016 BP 00002) and the program Ayudas Fundación BBVA (2017).
Acknowledgments
We would like to thank Antonio Forcina for making his Matlab code from [9] available to us. We have also benefited from discussions with Béatrice de Tilière.
Citation
Steffen Lauritzen. Caroline Uhler. Piotr Zwiernik. "Total positivity in exponential families with application to binary variables." Ann. Statist. 49 (3) 1436 - 1459, June 2021. https://doi.org/10.1214/20-AOS2007
Information