Preview

PDF BBOAJ...pdf | Download (355kB) | Preview

Abstract

Fitting a log binomial model to binary outcome data makes it possible to estimate risk and relative risk for follow-up data, and prevalenceand prevalence ratios for cross-sectional data. However, the fitting algorithm may fail to converge when the maximum likelihood solution is onthe boundary of the allowable parameter space. Some authorities recommend switching to Poisson regression with robust standard errors toapproximate the coefficients of the log binomial model in those circumstances. This solves the problem of non-convergence, but results in errorsin the coefficient estimates that may be substantial particularly when the maximum fitted value is large. The paradox is that the circumstancesin which the modified Poisson approach is needed to overcome estimation problems are the same circumstances when the error in using it isgreatest. We recommend that practitioners should be wary of using modified Poisson regression to approximate risk and relative risk.

Item Type:	Article
Authors/Creators:	Zhu, C and Blizzard, L and Stankovich, J and Wills, K and Hosmer, DW
Keywords:	relative risk, log binomial model, Poisson regression, boundary point
Journal or Publication Title:	Biostatistics and Biometrics Open Access Journal
Publisher:	Juniper Publishers
ISSN:	2573-2633
DOI / ID Number:	https://doi.org/10.19080/BBOAJ.2018.04.555649
Copyright Information:	Copyright © All rights are reserved by Blizzard L. Licensed under Creative Commons Attribution 4.0 International (CC BY 4.0) http://creativecommons.org/licenses/by/4.0/
Related URLs:	UTAS Profile: Blizzard, L Google Scholar: Stankovich, J UTAS Profile: Stankovich, J Google Scholar: Wills, K UTAS Profile: Wills, K
Item Statistics:	View statistics for this item

Actions (login required)

Item Control Page