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

1. Explain the underlying assumptions of a stationary
   4-dependent correlation structure as it pertains to
   the Infant Care Study.


2. Estimate the odds ratio and 95% confidence interval
   for the variable DIARRHEA (1 vs. 0) on a low weight-
   for-heightz-score (i.e., outcome¼1). Compute the
   95% confidence interval in two ways: first using the
   empirical standard errors and then using the model-
   based standard errors.


3. Referring to Exercise 2: Explain the circumstances in
   which the model-based variance estimators yield con-
   sistent estimates.


4. Referring again to Exercise 2: Which estimate of the
   95% confidence interval do you prefer?
   The following output should be used for Exercises 5–10
   and contains the results from running the same GEE
   model on the Infant Care data as in the previous questions,
   except that in this case, a stationary 8-dependent correla-
   tion structure is specified. The working correlation matrix
   for this model is included in the output.

Variable Coefficient

Empirical

Std Err

INTERCEPT 1.4430 1.2084

BIRTHWGT 0.0005 0.0003

GENDER 0.0014 0.5418

DIARRHEA 0.3601 0.8122

Stationary 8-Dependent Working Correlation Matrix

COL1 COL2 COL3 COL4 COL5 COL6 COL7 COL8 COL9

ROW1 1.0000 0.5255 0.3951 0.4367 0.4851 0.3514 0.3507 0.4346 0.5408

ROW2 0.5255 1.0000 0.5255 0.3951 0.4367 0.4851 0.3514 0.3507 0.4346

ROW3 0.3951 0.5255 1.0000 0.5255 0.3951 0.4367 0.4851 0.3514 0.3507

ROW4 0.4367 0.3951 0.5255 1.0000 0.5255 0.3951 0.4367 0.4851 0.3514

ROW5 0.4851 0.4367 0.3951 0.5255 1.0000 0.5255 0.3951 0.4367 0.4851

ROW6 0.3514 0.4851 0.4367 0.3951 0.5255 1.0000 0.5255 0.3951 0.4367

ROW7 0.3507 0.3514 0.4851 0.4367 0.3951 0.5255 1.0000 0.5255 0.3951

ROW8 0.4346 0.3507 0.3514 0.4851 0.4367 0.3951 0.5255 1.0000 0.5255

ROW9 0.5408 0.4346 0.3507 0.3514 0.4851 0.4367 0.3951 0.5255 1.0000

5. Compare the underlying assumptions of the stationary
   8-dependent correlation structure with the unstruc-
   tured correlation structure as it pertains to this model.


6. For the Infant Care data, how many more correlation
   parameters would be included in a model that uses an
   unstructured correlation structure rather than a sta-
   tionary 8-dependent correlation structure?

560 15. GEE Examples