# Mixed linear model with repeated measures in SPSS

I'm trying to run a mixed linear model analysis with SPSS. When I try to run the analysis, SPSS gives the following error message: The levels of the repeated effect are not different for each observation within a repeated subject.

I am trying to use a mixed model because my data has some missing values and the GLM does not work (almost every participant has at least one missing value).

The test used a within subject design and this is how my data looks like:

Participant | Repetition | Axis | X | Y | Z | P_rate

participant is the ID of the test subject. Repetition represents the repeated observation, while Axis, X Y Z represent the particular combination of conditions for that observation (Axis has 3 levels, X,Y,Z have two each). There were two repetitions for each combination. So I have a total of 48 observations (2 x 3 x 2 x 2 x 2). P_rate is the dependent variable.

In the dialog window I use Participant as my subject variable, Repetition as my repeated variable and Axis, X, Y, Z as my factors. However SPSS does not run the analysis because of that error.

What should I do in order for it to run the analysis correctly? I tried using an "Expectation Maximization" procedure to replace the missing values so that I could run a standard repeated measures ANOVA. It gives a significant difference for the Repetition parameter but just barely (p=0.049) so I wanted to use a mixed model to double check it.

Thanks in advance.

• Surely it's way too late for an answer (you posted this 1.5 yrs ago), but there is not /repeated line in this case. What you put in the /repeated line should have gone in the /random line. Commented Jan 28, 2014 at 4:48

## 1 Answer

This is an old post, but I'll try to answer it anyway for other people who see this. This message means that there is at least one participant who has some double value for repetition (within each participant, the values for repetition must be unique, for example 1,2,3 while your data might have 1,1,2 for some participant). So the problem is most likely in your data and not in the settings of the analysis.