Lab Home | Phone | Search
Center for Nonlinear Studies  Center for Nonlinear Studies
 Colloquia Archive 
 Postdoc Seminars Archive 
 Quantum Lunch 
 CMS Colloquia 
 Q-Mat Seminars 
 Q-Mat Seminars Archive 
 Kac Lectures 
 Dist. Quant. Lecture 
 Ulam Scholar 
 Summer Research 
 Student Application 
 Past Visitors 
 PD Travel Request 
 History of CNLS 
 Maps, Directions 
 CNLS Office 
Wednesday, July 24, 2019
10:00 AM - 11:00 AM
CNLS Conference Room (TA-3, Bldg 1690)


A Finite Selection Model (FSM) for Experimental Design

Carl Morris
Harvard University

The Health Insurance Experiment (HIE) was designed at RAND in the 1970s as one of the large-scale national public policy social experiments that could help to improve legislation. The HIE examined how different levels and types of health insurance proposals could impact health care expenditures and health status. In those years, we introduced the FSM as a new method for randomly assigning families to about 13 distinct health insurance plans (treatments). Relative to stratified random sampling, the FSM was seen back then to produce noticeably better balanced and more efficient samples for the treatments, with respect to the joint distributions of numerous (over 20) pre-experimental covariates that predict health utilization. The seminar at LANL will describe FSM's concepts and techniques that could deserve wider application in real experimental designs, and it compares and illustrates FSM's performance in relatively simple settings with certain other widely-used procedures. The work is part of an ongoing project at Harvard, jointly with Ambarish Chattopadhyay, PhD student, Statistics Department, and Assistant Professor Jose Zubizarreta, Health Care Policy Department.

Host: Sarah Michalak