Practice
Exercises
Questions 1–5 pertain to identifying the following correla-
tion structures that apply to clusters of four responses each:
A
10 :27 0:27 0: 27
0 :27 1 0:27 0: 27
0 :27 0:27 1 0:: 27
0 :27 0:27 0:27 12 6 6 6 6 6 4
3 7 7 7 7 7 5
B
10 :35 0 0
0 :35 1 0:35 0
00 :35 1 0: 35
000 :35 1
2 6 6 6 6 6 4
3 7 7 7 7 7 5 C
1000
0100
0010
0001
2
6
6
6
6
(^64)
3
7
7
7
7
(^75)
D
10 :50 0:25 0: 125
0 :50 1 0:50 0: 25
0 :25 0:50 1 0: 50
0 :125 0:25 0:50 1
2
6
6
6
6
(^64)
3
7
7
7
7
(^75)
E
10 :50 0:25 0: 125
0 :50 1 0:31 0: 46
0 :25 0:31 1 0: 163
0 :125 0:46 0:163 1
2
(^66)
6
6
6
4
3
(^77)
7
7
7
5
- Matrix A is an example of which correlation structure?
- Matrix B is an example of which correlation structure?
- Matrix C is an example of which correlation structure?
- Matrix D is an example of which correlation structure?
- Matrix E is an example of which correlation structure?
True or False (Circle T or F)T F 6. If there are two responses for each cluster, then
the exchangeable, AR1, and unstructured work-
ing correlation structure reduce to the same
correlation structure.
T F 7. A likelihood ratio test can test the statistical sig-
nificance of several parameters simultaneously
in a GEE model.
T F 8. Since GEE models produce consistent estimates
for the regression parameters even if the correla-
tion structure is misspecified (assuming the
mean response is modeled correctly), there
is no particular advantage in specifying the cor-
relation structure correctly.
T F 9. Maximum likelihood estimates are obtained in
a GLM by solving a system of score equations.536 14. Logistic Regression for Correlated Data: GEE