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

Test Questions 1–6 refer to models run on the data from the
Heartburn Relief Study (discussed in Sect. IV). In that
study, 40 subjects were given two symptom-provoking
meals spaced a week apart. Each subject was administered
an active treatment following one of the meals and a stan-
dard treatment following the other meal, in random order.
The goal of the study was to compare the effects of an
active treatment for heartburn with a standard treatment.
The dichotomous outcome is relief from heartburn (coded
1 ¼yes, 0¼no). The exposure of interest is RX (coded
1 ¼active treatment, 0¼standard treatment). Addition-
ally, it was hypothesized that the sequence in which each
subject received the active and standard treatment could
be related to the outcome. Moreover, it was speculated that
the treatment sequence could be an effect modifier for the
association between the treatment and heartburn relief.
Consequently, two other variables are considered for the
analysis: a dichotomous variable SEQUENCE and the
product term RX*SEQ (RX times SEQUENCE). The vari-
able SEQUENCE is coded 1 for subjects in which the active
treatment was administered first and 0 for subjects in
which the standard treatment was administered first.

The following printout summarizes the computer output

for three GEE models run on the heartburn relief data

(Model 1, Model 2, and Model 3). An exchangeable corre-

lation structure is specified for each of these models. The

variance–covariance matrix for the parameter estimates

and the Score test for the variable RX*SEQ are included

in the output for Model 1.

Model 1

Variable Coefficient

Empirical

Std Err

INTERCEPT 0.6190 0.4688

RX 0.4184 0.5885

SEQUENCE 0.8197 0.6495

RX*SEQ 0.2136 0.7993

Empirical Variance Covariance Matrix

For Parameter Estimates

INTERCEPT RX SEQUENCE RX*SEQ

INTERCEPT 0.2198 0.1820 0.2198 0.1820

RX 0.1820 0.3463 0.1820 0.3463

SEQUENCE 0.2198 0.1820 0.4218 0.3251

RX*SEQ 0.1820 0.3463 0.3251 0.6388

Score test statistic for RX*SEQ¼0.07

562 15. GEE Examples