Logistic Regression: A Self-learning Text, Third Edition (Statistics in the Health Sciences)

Presentation

I. Overview

Examples:

Modeling

outcomes with

dichotomous

correlated

responses

FOCUS

Three examples are presented:

1. Infant Care Study


2. Aspirin–Heart Bypass Study


3. Heartburn Relief Study

II. Example 1: Infant Care
Study

Introduced in Chap. 14

Response (D):

Weight-for-height standardized (z)
score

D¼

1 if z<---1 (‘‘Wasting’’)

0 otherwise

(

In this chapter, we provide examples of how

the GEE approach is used to carry out logis-

tic regression for correlated dichotomous

responses.

We examine a variety of GEE models using

three databases obtained from the following

studies: (1) Infant Care Study, (2) Aspirin–

Heart Bypass Study, and (3) Heartburn Relief

Study.

In Chap. 14, we compared model output from

two models run on data obtained from an

infant care health intervention study in Brazil

(Cannon et al., 2001). We continue to examine

model output using these data, comparing

the results of specifying different correlation

structures.

Recall that the outcome of interest is a dicho-

tomous variable derived from a weight-

for-height standardized score (i.e., z-score)

obtained from the weight-for-height distribu-

tion of a reference population. The dichoto-

mous outcome, an indication of “wasting,” is

coded 1 if thez-score is less than negative 1,

and 0 otherwise.

542 15. GEE Examples