I am about to start data analysis of a small randomized controlled trial (n = 60) with two groups and three time points (baseline, follow up1 and follow-up 2). I chose to run an LMM to account for the dependency between the repeated measures across time points.
"subject ID" is the hierarchical variable;
Fixed effect: Group, Time and baseline leg pain intensity (primary outcome set as covariate) are fixed effects. The main effect of the covariate and group by time interactions were in fixed effects.
Random effect: main effect of time with an "AR(1): heterogenous" covariance structure;
Statistics: Parameter estimates and Tests for covariance parameters
Estimated Marginal Means: Group*Time interactions and pairwise comparisons
The group*time output is shown in the attached file. The MIXED procedure does not provide between-groups mean differences and its respective 95% confidence intervals, but I need this data. Do I need to manually calculate it?
You can manually add the line
/EMMEANS=TABLES(Group*Time) COMPARE(Group)
to SPSS command syntax in order to obtain pairwise group comparisons.
Assuming you use deviation coding (not dummy coding), then the model should include a coefficient for 'Grupo' which corresponds to the main effect.
Hi Giovanni
We recently discovered a problem with LMM in SPSS v23/v24. Can you fit a GLM ignoring repeated measures and check the results (beta coefficients and standard errors) and compare with LMM. For our data these two results are the same, hence a bug in LMM since v23. SPSS has introduced a GLMM, and this works fine. We are trying to find out what has happened. Please try this and let me know what you get.
Thanks
Max
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS