Hi!
I am trying to understand how personality traits change in adolescents over time (2 measurement occasions with the same individuals). I am using linear mixed models in SPSS for that purpose. I have "subjects" in "subjects" in the first block, "extraversion" as a DV, and "time" as a covariate in the second block (time could be a factor as well but I read that it might be better to put it in the covariate block actually). For fixed effects, I choose time as this is my main interest.
My question is if I should put "time" also in the first block under "repeated".
The analysis is very limited due to the fact that there are only 2 measurement occasions, so I cannot lose many degrees of freedom. If I do not add time in the repeated block, I can still add a random intercept and slope, but if I have to add time also under repeated, then I hit easily the iteration warning. Furthermore, I could add "sex" as a covariate, but to make all of these decisions, I first need to know if, for this particular design, where I only have time as a predictor, I need to have it in both blocks (repeated measures & covariate)?
I think that I do need to add it in both boxes. In that scenario I am not able to add any random effects due to iteration, but I can still add covariate of sex and interaction of sex and time. Please let me know if this is correct!
Thanks a lot!
Best wishes
Michaela
Hi,
The repeated measure is "time" on a DV. Covariate effect may or may not be present.No need to add again time in covariate.
Hi, I need to add "time" as a factor (or covariate), because I am interested in the fixed effect of time, as I want to see a development (or change) on the DV, I cannot do this if I put "time" only in repeated as I only get output on the intercept.
With two time points there are other methods that are often used rather than the multilevel approach. Google "Lord's paradox" for some of this discussion.
If with "other" you mean designs such as AN(C)OVA, they are quite disadvantaged with respect to handling missing data and assumptions.
Hi everyone I have found out that there are two ways of accounting for the between subject variance and I can choose either random effects OR adjusting for the covariance. Not both :)
- Badges
- Science topic
- Similar topics
- Mathematical Sciences
- Statistics
- Statistical Software
- SPSS