MANUAL Published: 01 January 2009
MNL11489M

Appendix H: A Proportional Odds/Hazards Approach to JAR Data

Source

t-tests, ANOVA, and linear regression all assume that the response is measured on an interval scale, so that the differences between adjacent values have the same meaning across the scale. This assumption is often violated in practice, which can lead to inaccurate conclusions. The proportional odds and proportional hazards models are ordinal regression models that are only sensitive to the order of the observations, not the specific values assigned to the categories. They are used to compare the distributions of JAR scores among products and are performed simultaneously. The proportional odds model (POM) is skew-symmetric, so that reversing the order of the scale simply changes the sign of the mean, while the proportional hazards model (PHM) is asymmetric, so that reversing the order of the scale changes both the order and the sign of the estimate.

Author Information

Xiong, Rui
University of Arkansas, Fayetteville, AR
Meullenet, Jean-Francois
University of Arkansas, Fayetteville, AR
Price: $25.00
Contact Sales
Related
Reprints and Permissions
Reprints and copyright permissions can be requested through the
Copyright Clearance Center
Details
Developed by Committee: E18
Pages: 38–41
DOI: 10.1520/MNL11489M
ISBN-EB: 978-0-8031-6739-1
ISBN-13: 978-0-8031-7010-0