LMATRIX in MIXED procedure
Locked
 
|
Hello,
A question:
I have a 2x7 (group by time) design. I need to obtain 6 single df interaction
contrasts (group by time1 vs. time1+j, where j ranges from 1 to 6).
I know how to set this up using CONTRAST subcommand in MANOVA
procedure (see below) and I realize that the standard errors are
not correct. The dv=LGWBIFN, group=TPO, and time=VISIT.
MANOVA LGWBIFN BY TPO(0,1) VISIT(1,7)
/PRINT=CELLINFO(MEANS) SIGNIF(EFSIZE)
/CONTRAST(VISIT)=SIMPLE(1)
/DESIGN=TPO VISIT(1) VISIT(2) VISIT(3) VISIT(4) VISIT(5) VISIT(6)
TPO*VISIT(1) TPO*VISIT(2) TPO*VISIT(3) TPO*VISIT(4) TPO*VISIT(5) TPO*VISIT(6).
What I am attempting to do is analyze the data correctly using the MIXED procedure
(due in part to unequal number of time points among subjects) but I cannot
seem to figure out how to set up the LMATRIX to obtain the correct tests.
Below is an INcorrect attempt.
MIXED LGWBIFN BY VISIT TPO
/FIXED=TPO VISIT TPO*VISIT
/TEST = 'main effects & contrasts'
ALL 1 1 1 1 1 1 1 -1 -1 -1 -1 -1 -1 -1;
ALL -1 1 0 0 0 0 0 -1 1 0 0 0 0 0;
ALL -1 0 1 0 0 0 0 -1 0 1 0 0 0 0;
ALL -1 0 0 1 0 0 0 -1 0 0 1 0 0 0;
ALL -1 0 0 0 1 0 0 -1 0 0 0 1 0 0;
ALL -1 0 0 0 0 1 0 -1 0 0 0 0 1 0;
ALL -1 0 0 0 0 0 1 -1 0 0 0 0 0 1;
ALL -1 1 0 0 0 0 0 1 -1 0 0 0 0 0;
ALL -1 0 1 0 0 0 0 1 0 -1 0 0 0 0;
ALL -1 0 0 1 0 0 0 1 0 0 -1 0 0 0;
ALL -1 0 0 0 1 0 0 1 0 0 0 -1 0 0;
ALL -1 0 0 0 0 1 0 1 0 0 0 0 -1 0;
ALL -1 0 0 0 0 0 1 1 0 0 0 0 0 -1
/METHOD=REML
/PRINT=LMATRIX SOLUTION
/RANDOM=INTERCEPT | SUBJECT(ID) COVTYPE(VC).
Although this design matrix is what I would do in MANOVA,
MIXED wants something else. Can anyone give me some help?
Regards,
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
|
|
|
Jason,
Before responding to your question, I'd like to suggest that you consider including a REPEATED statement with a residual variance-covariance matrix that best fit your data. Some options that come to mind include an autoregressive matrix or perhaps even an unstructured matrix. If the within-subject correlation of error terms is roughly equal, then the RANDOM intercept statement by itself would probably suffice. Much can be said on this topic, but I'll move on to your question. I am not entirely clear on what you're asking, but to help get you started, if you wanted to contrast the groups on the difference in mean change from the first time point to the seventh time point, you could write the following subcommand code: /TEST = 'Contrast' group*time 1 0 0 0 0 0 -1 -1 0 0 0 0 0 1 If you were interested in contrasting the groups on the difference in mean change from the first time point to the sixth time point, you could write the following subcommand code: /TEST = 'Contrast' group*time 1 0 0 0 0 -1 0 -1 0 0 0 0 1 0 I have not tested this code, but you should able to confirm that no errors have been made quite easily. One final point--There are many assumptions to run a linear mixed model that you consider investigating before running this type of analysis. HTH, Ryan p.s. Feel free to write back if I misunderstood your design/question or if you find errors in the code I provided.
|
«
Return to SPSSX Discussion
|
1 view|%1 views
| Free forum by Nabble | Edit this page |