factorial repeated measures in GLM vs MIXED procedures.

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

factorial repeated measures in GLM vs MIXED procedures.

Beckstead, Jason

Hellow Listers,

 

 

I have created 64 unique patient descriptions using a 2x2x2x2x2x2 repeated measures

factorial design.

Each cell in the factorial design describes a patient in terms of being

older or younger, male or female, having had bypass surgery or not,

having diabetes or not, having muscular-skeletal pain or not,

and having high motivation to exercise or not.

 

Physicians read each description and then provide a judgment (0 to 100 scale)

of how likely it is that the patient will BENEFIT from cardiac rehabilitation.

 

For now, I am interested only in main effects and 2-way interactions involving FEMALE.

 

I successfully analyze the data as repeated measures (wide format) using GLM

 procedure as follows:

 

GLM BENEFIT_1 TO  BENEFIT_64

 /WSFACTOR=OLDER(2) FEMALE (2) BYPASS (2) DIAB (2) MSPAIN (2) HIMOT (2)

 /WSDESIGN=OLDER FEMALE BYPASS DIAB MSPAIN HIMOT

  FEMALE*OLDER FEMALE*BYPASS FEMALE*DIAB FEMALE*MSPAIN FEMALE*HIMOT.

 

I am trying to duplicate this analysis with MIXED procedure for illustration purposes.

The file has been restructured so that there are 64 records per case.

 

I use the following syntax but it does not give correct tests (degrees of freedom

 for error terms seem to be based on total # of observations, rather than on # subjects).

 

MIXED BENEFIT BY OLDER FEMALE BYPASS DIAB MSPAIN HIMOT

 /FIXED= OLDER FEMALE BYPASS DIAB MSPAIN HIMOT

  FEMALE*OLDER FEMALE*BYPASS FEMALE*DIAB FEMALE*MSPAIN FEMALE*HIMOT

 /METHOD=REML

 /PRINT= TESTCOV

 /REPEATED=OLDER FEMALE BYPASS DIAB MSPAIN HIMOT

  FEMALE*OLDER FEMALE*BYPASS FEMALE*DIAB FEMALE*MSPAIN FEMALE*HIMOT | SUBJECT(SUB) COVTYPE(CS).

 

I am not very familiar with MIXED syntax. Can someone offer corrections/suggestions?

 

Thanks.

Jason

_____________________________________________________________

 Jason W. Beckstead, Ph.D.       

  Associate Professor/Quantitative Methodologist

  University of South Florida College of Nursing

  12901 Bruce B. Downs Blvd., MDC22, Tampa, FL 33612, USA

  Statistical Editor, International Journal of Nursing Studies

  phone: +1.813.974.7667  fax: +1.813.974.5418          

  personal website:  http://personal.health.usf.edu/jbeckste/

  International Journal of Nursing Studies  http://www.elsevier.com/ijns

 

 

 

Reply | Threaded
Open this post in threaded view
|

Re: factorial repeated measures in GLM vs MIXED procedures.

Ryan
Jason,

Please provide an illustration of your dataset in vertical format, starting with the patient identification variable. Further, please explain how you have repeated measures. 

Ryan

Sent from my iPhone

On Mar 25, 2012, at 1:51 PM, "Beckstead, Jason" <[hidden email]> wrote:

Hellow Listers,

 

 

I have created 64 unique patient descriptions using a 2x2x2x2x2x2 repeated measures

factorial design.

Each cell in the factorial design describes a patient in terms of being

older or younger, male or female, having had bypass surgery or not,

having diabetes or not, having muscular-skeletal pain or not,

and having high motivation to exercise or not.

 

Physicians read each description and then provide a judgment (0 to 100 scale)

of how likely it is that the patient will BENEFIT from cardiac rehabilitation.

 

For now, I am interested only in main effects and 2-way interactions involving FEMALE.

 

I successfully analyze the data as repeated measures (wide format) using GLM

 procedure as follows:

 

GLM BENEFIT_1 TO  BENEFIT_64

 /WSFACTOR=OLDER(2) FEMALE (2) BYPASS (2) DIAB (2) MSPAIN (2) HIMOT (2)

 /WSDESIGN=OLDER FEMALE BYPASS DIAB MSPAIN HIMOT

  FEMALE*OLDER FEMALE*BYPASS FEMALE*DIAB FEMALE*MSPAIN FEMALE*HIMOT.

 

I am trying to duplicate this analysis with MIXED procedure for illustration purposes.

The file has been restructured so that there are 64 records per case.

 

I use the following syntax but it does not give correct tests (degrees of freedom

 for error terms seem to be based on total # of observations, rather than on # subjects).

 

MIXED BENEFIT BY OLDER FEMALE BYPASS DIAB MSPAIN HIMOT

 /FIXED= OLDER FEMALE BYPASS DIAB MSPAIN HIMOT

  FEMALE*OLDER FEMALE*BYPASS FEMALE*DIAB FEMALE*MSPAIN FEMALE*HIMOT

 /METHOD=REML

 /PRINT= TESTCOV

 /REPEATED=OLDER FEMALE BYPASS DIAB MSPAIN HIMOT

  FEMALE*OLDER FEMALE*BYPASS FEMALE*DIAB FEMALE*MSPAIN FEMALE*HIMOT | SUBJECT(SUB) COVTYPE(CS).

 

I am not very familiar with MIXED syntax. Can someone offer corrections/suggestions?

 

Thanks.

Jason

_____________________________________________________________

 Jason W. Beckstead, Ph.D.       

  Associate Professor/Quantitative Methodologist

  University of South Florida College of Nursing

  12901 Bruce B. Downs Blvd., MDC22, Tampa, FL 33612, USA

  Statistical Editor, International Journal of Nursing Studies

  phone: +1.813.974.7667  fax: +1.813.974.5418          

  personal website:  http://personal.health.usf.edu/jbeckste/

  International Journal of Nursing Studies  http://www.elsevier.com/ijns