I haven't been able to work out whether I should use a Generalised Linear Mixed Model (GLMM) or a Linear Mixed Model (LMM) for my analysis (2 time points with both continuous & categorical IVs. Continuous & categorical DVs. Ie, PTSD Sx score & PTSD Dx). From what I can gather, the former can be used for normal & non-normal distributions for the latter only for normal distributions. Any thoughts would be appreciated.
Use GLMM if you have missing data at either time point to prevent deleting cases.
Thanks Letoynia. The other problem I was having was that the GLMM was deleting cases (e.g. 164 of 256 long format cases) based on missing variables. The whole point of performing a GLMM rather than mixed ANOVA/SPANOVA was to minimise excluded cases based on missing variables. Might you know why it was doing this and how to fix it?
You would need to do your analysis in Mplus (or possibly some other SEM software) which uses all available data by default or impute the missing data.
By the way, I realized later that your 'G' meant generalized and not general. So, in response to your first question without actually seeing your data, I would use general-lmm for your continuous outcomes and generalized-lmm for your categorical outcomes.
I will have to work out how to use all the available data/imputation in SPSS unfortunately. I am using Generalised LMM for all analyses, as the continuous outcomes are non-normally distributed. Thanks Letoynia.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Applied Statistics