Hi-- I am using the SPSS MIXED analysis for a clinical trial (2 treatment groups X 2 time points). Because there are missing data points, I'd like to be able to provide the intent-to-treat means and standard deviations. The EMMEANS subcommand only provides one "grand" mean but does not allow me to break it down by treatment and time point. Any suggestions would be most appreciated.
Dear David,
To break the means between treatment and time point, they have to be included in the model as factors. For example, see the syntax below. The problem with this approach is that there is no difference between subjects and it assumes that time can be treated as categorical. Alternatively, if time is important, you can ask to save the predict values of the model.
MIXED IMP BY treatment period
/FIXED= treatment period treatment*period | SSTYPE(3)
/METHOD= REML
/PRINT= SOLUTION TESTCOV
/REPEATED=period | SUBJECT(patientnr) COVTYPE(UN)
/EMMEANS=TABLES(treatment* period) .
Best,
Thanks for your reply, Dario. I tried running this syntax, substituting my own variable names, but for some reason SPSS doesn't like it (several different error messages come up, depending on what I do). But I will keep playing with it. Best, David
You can achieve similar results by menu. Check variable types, data format etc.
Dear David,
Please find below the full syntax. I hope that can help you out further.
MIXED IMP BY treatment period
/CRITERIA= CIN(95) MXITER(100) MXSTEP(5) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE)
/FIXED= treatment period treatment*period | SSTYPE(3)
/METHOD= REML
/PRINT= SOLUTION TESTCOV
/REPEATED=period | SUBJECT(patientnr) COVTYPE(UN)
/EMMEANS=TABLES(treatment*period) .
Best,
Thank you Dario. Still happening, I'm afraid. My variables (all numeric format) are:
ID (patient ID #)
OUTCOME (the questionnaire we're administering pre- and post-study)
TREATMENT (1 = treated, 2 = not treated)
TIME (1 = beginning, 2 = end)
So modifying your syntax I used:
MIXED outcome BY treatment time
/CRITERIA= CIN(95) MXITER(100) MXSTEP(5) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE)
/FIXED= treatment time treatment*time | SSTYPE(3)
/METHOD= REML
/PRINT= SOLUTION TESTCOV
/REPEATED=time | SUBJECT(id) COVTYPE(UN)
/EMMEANS=TABLES(treatment*time) .
And the error message is:
Warnings
Text: Command: MIXED
An undefined variable name, or a scratch or system variable was specified in a variable list which accepts only standard variables. Check spelling and verify the existence of this variable.
Execution of this command stops.
Mehmet, thank you. Any specifics you can provide would be most appreciated.
Could one solution be to use the SAVE subcommand and, under "predicted values & residuals," select "predicted values?" Does that then provide the estimated numbers from which I could then run DESCRIPTIVES and use to describe the sample?
I don't have the solution. Did you search the error message on google?
Your suggestion should not work. Because the mixed command is not executed, you can't have a fit and residual. Did you make sure that your variable names are valid?
HI Mehmet-- yes I checked and double-checked the variables and that doesn't seem to be the problem... What is curious is that this command runs just fine (though I can't get the means):
MIXED outcome WITH time treatment
/CRITERIA=CIN(95) MXITER(100) MXSTEP(10) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0,
ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE)
/FIXED=time treatment time*treatment | SSTYPE(3)
/METHOD=REML
/PRINT=SOLUTION TESTCOV
/RANDOM=INTERCEPT | SUBJECT(ID) COVTYPE(UN)
/SAVE=PRED.
Dear David,
Sorry to hear that. Can you actually run the model withoutthe EMMEANS?
Is your data bank in the long format as well?
You can save the predicted values ( not residuals). But if you can run the model, it should be able to get the EMMEANS.
Best,
Dear David, when you use statement with the variables are covariates. Then you cannot request EMMEANS. Can you run the same model with the statement by?
Best
Thanks Dario. It seems like my original syntax works if I change "WITH" to "BY," specifically:
MIXED outcome by time treatment
/CRITERIA=CIN(95) MXITER(100) MXSTEP(10) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0,
ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE)
/FIXED=time treatment time*treatment | SSTYPE(3)
/METHOD=REML
/PRINT=SOLUTION TESTCOV
/RANDOM=INTERCEPT | SUBJECT(ID) COVTYPE(UN)
/EMMEANS=TABLES(treatment*time) .
SPSS output file is attached.
Does that look right to you?
Dear David,
Sorry for my late reaction. The output looks good. I'm glad that now you were able to run the model.
The warning message is not a problem as well.
Best,
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS