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

Presentation

I. Overview

Other approaches

to modeling

outcomes with

dichotomous

correlated

responses

FOCUS

Other approaches for correlated
data:

1. Alternating logistic regressions
   (ALR) algorithm


2. Conditional logistic regression


3. Generalized linear mixed
   model

In this chapter, we provide an introduction to

modeling techniques other than GEE for use

with dichotomous outcomes in which the

responses are correlated.

In addition to the GEE approach, there are a

number of alternative approaches that can be

applied to model correlated data. These

include (1) the alternating logistic regressions

algorithm, which uses odds ratios instead of

correlations, (2) conditional logistic regres-

sion, and (3) the generalized linear mixed

model approach, which allows for random

effects in addition to fixed effects. We briefly

describe each of these approaches.

This chapter is not intended to provide a thor-

ough exposition of these other approaches but

rather an overview, along with illustrative exam-

ples, of other ways to handle the problem of

analyzing correlated dichotomous responses.

Some of the concepts that are introduced in

this presentation are elaborated in the Practice

Exercises at the end of the chapter.

Conditional logistic regression has previously

been presented in Chap. 11 but is presented

here in a some-what different context. The

alternating logistic regression and generalized

linear mixed model approaches for analyzing

correlated dichotomous responses show great

promise but at this point have not been fully

investigated with regard to numerical estima-

tion and possible biases.

570 16. Other Approaches for Analysis of Correlated Data