Mixed Model Mishaps

[Byrd Sellers, Johnita]

Hello All,

 

I have posted a similar email to the one below but I received no
responses.  After learning a couple of things on my own, I thought that
I would be able to analyze the data but I am still running into a couple
of problems.  I added parts of the original email to help in
understanding what I am doing.  Any help regarding the below problem
would be great.  Also, does anyone know of a good course that I can take
on Mixed Models and repeated measures (in Georgia)?  

 

I am working with a problem that requires me to use GLM to solve, in
particular, mixed models.  I have tried analyzing the problem using SAS
and I would like to try analyzing it using SPSS.  I have been having a
little difficulty in setting up the problem in SAS.  In essence, this is
what I am trying to do:

I am trying to look at different pressure values of 20 patients using
two different types of devices.  I want to compare the two devices to
one another.  In other words, I would like to see what the values are on
one device, let's say device 2, when the values of device 1 is 0.  I
have set up the model as follows (I have supplied a portion of the model
so that you can have a better understanding):

Machine

Patient

Reading

1

1

67

1

1

69

1

1

63

1

1

62

1

1

61

1

1

59

1

1

56

1

1

55

1

1

31

1

1

16

1

1

15

2

1

71

2

1

65

2

1

56

2

1

36

2

1

26

2

1

17

2

1

5

2

1

-1

2

1

-11

2

1

-21

2

1

-35

1

2

65

1

2

66

1

2

64

1

2

63

1

2

57

1

2

58

1

2

60

1

2

60

1

2

58

1

2

53

1

2

28

1

2

20

2

2

88

2

2

75

2

2

63

2

2

58

2

2

48

2

2

43

2

2

35

2

2

15

2

2

4

2

2

-13

2

2

-25

2

2

-34

1

3

63

1

3

66

1

3

67

1

3

65

1

3

66

1

3

65

1

3

64

1

3

61

1

3

58

1

3

52

1

3

33

1

3

20

2

3

74

2

3

59

2

3

49

2

3

42

2

3

27

2

3

15

2

3

10

2

3

2

2

3

-11

2

3

-6

2

3

-28

2

3

-34

1

4

72

1

4

79

1

4

74

1

4

71

1

4

71

1

4

73

1

4

71

1

4

70

1

4

68

1

4

65

1

4

63

1

4

47

1

4

39

2

4

71

2

4

60

2

4

48

2

4

41

2

4

32

2

4

21

2

4

20

2

4

5

2

4

-9

2

4

-13

2

4

-26

2

4

-32

2

4

-46

1

5

69

1

5

70

1

5

72

1

5

69

1

5

69

1

5

66

1

5

66

1

5

64

1

5

62

1

5

59

1

5

35

Where machine represents the two devices, patient represents the
different patients and reading represents the pressure reading that was
displayed on the devices.

Here is how I have decided to set this problem up in SPSS, I let the
SUBJECT be the patients, DEPENDENT VARIABLE is the readings, and the
FACTOR is the machine.  I have set up the machine to be a fixed effect
and I have selected the patients to be a random effect.  I am having a
problem deciding what Covariance Structure should be.  I first decided
to use Variance Components but I am getting a lot of blank information.
Also, I analyzed the data using both SPSS and SAS with my covariance
structure being Variance Components and I am not coming up with the same
thing.  If there is anyone who can help me or tell me what I am doing
wrong, I would appreciate it.  (Each patient has repeated measures with
each one of the devices.  Also, each patient uses BOTH of the devices.)


 

Johnita