Troubleshooting
Problem
I am running a GLM - Repeated Measures with a Sum of squares type IV. I have observed that it only retains cases without missing data, and there are few of them - most of them have incomplete data. My question is : which model to use in order to take into account all the data, even with cases that have missing values?
Resolving The Problem
There is no option to get GLM to use more cases with repeated measures. It uses the multivariate general linear model approach, which requires complete data for all cases. You may want to consider using a mixed models approach, which would require restructuring the data. The following commands do this and also recode the index variable created (which I've called time, but she could call it something different if desired) to index repeated measures so that the values match the numeric pieces of the original dependent variable names.
VARSTOCASES
/MAKE volume FROM vol9 vol14 vol22 vol29 vol36 vol41 vol48 vol55 volL64
/INDEX=time(9)
/KEEP=souris ttt
/NULL=DROP.
RECODE time (1=9) (2=14) (3=22) (4=29) (5=36) (6=41) (7=48) (8=55) (9=64).
EXECUTE.
The recoding is done based on the assumption that there's a metric for time that's not equally spaced, and that you might want to treat time as a covariate and fit only a linear effect or perhaps linear and quadratic, or whatever, which you can do in MIXED. You may want to get some statistical help if you are not familiar with mixed models approaches, as these data look problematic in that it appears that there's both correlation among repeated measurements and inequality in variances over time, which would indicate the need to fit a covariance structure for the R matrix that allows for both, but the lack of data at later time points seems to be making this problematic.
Related Information
Historical Number
80253
Was this topic helpful?
Document Information
Modified date:
16 April 2020
UID
swg21482394