Abstract Details
Activity Number:
|
464
|
Type:
|
Contributed
|
Date/Time:
|
Wednesday, August 6, 2014 : 8:30 AM to 10:20 AM
|
Sponsor:
|
Section on Statistical Computing
|
Abstract #311424
|
View Presentation
|
Title:
|
MM Optimization in Massive Observational Analysis
|
Author(s):
|
Trevor R. Shaddox*+ and Kenneth L. Lange and David Madigan and Marc Suchard
|
Companies:
|
University of California, Los Angeles and University of California, Los Angeles and Columbia University and University of California, Los Angeles
|
Keywords:
|
MM algorithm ;
pharmacovigilance ;
Bayesian ;
parallelization
|
Abstract:
|
Adverse drug events (ADEs) remain a serious public health risk. Identifying dangerous drugs from the emerging national patient claims and electronic health record databases is a non-trivial statistical challenge. Specifically, fitting models in the context of datasets with tens of thousands of covariates and millions of observations always perches on the edge between computationally expensive and intractable. New techniques for optimization in this setting add to the arsenal of strategies that can push sophisticated, meaningful modeling toward feasiblity. Here, we incorporate ideas from the Majorization-Minimization and Minorization-Maximization (MM) approach to develop two novel MM algorithms in the context of the Bayesian self-controlled case series (BSCCS) regression model. In the first algorithm, we take two minorization transformations of the BSCCS likelihood and optimize with sequential Newton steps. In the second algorithm, we take a single minorization transformation and solve the resulting minorizing quadratics. The result of both techniques is parameter separation in the surrogate at each optimization step, at the cost of an increase in the number of iterations requir
|
Authors who are presenting talks have a * after their name.
Back to the full JSM 2014 program
|
2014 JSM Online Program Home
For information, contact jsm@amstat.org or phone (888) 231-3473.
If you have questions about the Professional Development program, please contact the Education Department.
The views expressed here are those of the individual authors and not necessarily those of the JSM sponsors, their officers, or their staff.
Copyright © American Statistical Association.