|
Activity Number:
|
538
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 2, 2007 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Biopharmaceutical Section
|
| Abstract - #309250 |
|
Title:
|
Regression Models for Multinomial Responses with an Example from PCPT
|
|
Author(s):
|
Kenneth Liu*+ and Alexandra Carides
|
|
Companies:
|
Merck & Co., Inc. and Merck & Co., Inc.
|
|
Address:
|
UG1C46, North Wales, PA, 19454-1099,
|
|
Keywords:
|
Polytomous logistic regression ; nominal response ; ordinal response ; proportional odds assumption
|
|
Abstract:
|
Logistic regression is used to analyze a binary response variable. When the response variable has more than two levels, various forms of multinomial logistic regression can be used. We will discuss two types that permit non-proportional odds. Nominal logistic regression is a direct extension of logistic regression where each unordered response category is compared to a common reference. The unconstrained partial proportional odds model is one type of ordinal logistic regression that models an ordered response using cumulative logits. In the Prostate Cancer Prevention Trial (PCPT), patients can have high-grade, ungraded, low-grade, or no cancer. Nominal logistic regression compares each grade of cancer to a reference group such as the no cancer category. The unconstrained partial proportional odds model orders the cancer categories and compares cumulative categories.
|
- 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 2007 program |
