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

Test True or false (Circle T or F)
T F 1. A model with subject-specific random effects is
an example of a marginal model.
T F 2. A conditional logistic regression cannot be used
to obtain parameter estimates for a predictor
variable that does not vary its values within the
matched cluster.
T F 3. The alternating logistic regressions approach
models relationships between pairs of responses
from the same cluster with odds ratio para-
meters rather than with correlation parameters
as with GEE.
T F 4. A mixed logistic model is a generalization of the
generalized linear mixed model in which a link
function can be specified for the modeling of the
mean.
T F 5. For a GEE model, the user specifies a correlation
structure for the response variable, whereas for a
GLMM, the user specifies a covariance structure.

Questions 6–10 refer to models run on the data from the

Heartburn Relief Study. The following printout sum-

marizes the computer output for two mixed logistic models.

The models include a subject-specific random effect for the

intercept. The dichotomous outcome is relief from heart-

burn (coded 1¼yes, 0¼no). The exposure of interest is RX

(coded 1¼active treatment, 0¼standard treatment). The

variable 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

product term RX*SEQ (RX times SEQUENCE) is included

to assess interaction between RX and SEQUENCE. Only

the estimates of the fixed effects are displayed in the output.

Model 1

Variable Estimate Std Err

INTERCEPT 0.6884 0.5187

RX 0.4707 0.6608

SEQUENCE 0.9092 0.7238

RX*SEQ 0.2371 0.9038

Model 2

Variable Coefficient Std Err

INTERCEPT 0.6321 0.4530

RX 0.3553 0.4565

SEQUENCE 0.7961 0.564

Test 595