JSM 2004 - Toronto

Abstract #301273

This is the preliminary program for the 2004 Joint Statistical Meetings in Toronto, Canada. Currently included in this program is the "technical" program, schedule of invited, topic contributed, regular contributed and poster sessions; Continuing Education courses (August 7-10, 2004); and Committee and Business Meetings. This on-line program will be updated frequently to reflect the most current revisions.

To View the Program:
You may choose to view all activities of the program or just parts of it at any one time. All activities are arranged by date and time.

The views expressed here are those of the individual authors
and not necessarily those of the ASA or its board, officers, or staff.

Back to main JSM 2004 Program page
Activity Number: 382
Type: Contributed
Date/Time: Wednesday, August 11, 2004 : 2:00 PM to 3:50 PM
Sponsor: Section on Statistics in Epidemiology
Abstract - #301273
Title: Limitations of the Log Binomial Regression Model for Estimating Risk or Prevalence Ratios
Author(s): Leigh Blizzard*+ and David W. Hosmer
Companies: University of Tasmania and University of Massachusetts
Address: Menzies Research Instiute, Hobart Australia 7008, International, 7008, Australia
Keywords: binary regression ; odds ratio ; logistic regression
Abstract:

An estimate of the risk or prevalence ratio adjusted for confounders can be obtained from logistic regression, but it substantially overestimates when the outcome is not rare. The log binomial model, binomial errors and log link, is increasingly being used for this purpose. Its performance and appropriately modified logistic regression fit tests have not been evaluated. Extensive simulations of artificial data and bootstrap sampling of real data are used to compare the performance of the log binomial, logistic regression and a logistic regression-based method proposed by Schouten et al. (1993). Log binomial regression resulted in "failure" rates (nonconvergence, out-of-bounds predicted probabilities) as high as 59%. Estimates from the Schouten method produced fitted log binomial probabilities greater than one in up to 19% of samples. Coefficient estimates from these two models were similar, but the Schouten method over estimated the standard errors. Rejection rates for the Hosmer-Lemeshow, Stukel goodness-of-link test and the unweighted sum of squares tests were around 5%. No test had high power. Unquestioned use of the log binomial model is not recommended.

  • The address information is for the authors that have a + after their name.
  • Authors who are presenting talks have a * after their name.

Back to the full JSM 2004 program

JSM 2004 For information, contact jsm@amstat.org or phone (888) 231-3473. If you have questions about the Continuing Education program, please contact the Education Department.
Revised March 2004