MANCOVA and MANOVA Discussion > Repeated measures with multiple factors/covariates and missing data!
Hi Sophia!
From what you explain in your post, I would recommend using a random coefficients model with R (a growth model, specifically). Paul Bliese wrote a great R package and wrote a wonderful article walking through each step of running the model. Best of all, R is completely free to use (though not super user-friendly).
A random coefficients growth model would be a good choice for you because it would allow you to use all the data that you have and not exclude participants with listwise deletion.
Here is a link to Dr. Bliese's article:
Paul Bliese Growth Modeling Article
April 11, 2013 |
Jeremy Taylor
Jeremy Taylor
Hi,
I'd be really grateful for any advice on how to proceed with this!
I've got subjects from 5 different groups, created by different combinations of two factors: location and day-length. Each subject completed a questionnaire at each of 3 time-points, to produce scores for 16 psychological measures, so they each should have 48 scores altogether. Most (but not all) of the 48 scores show an approximately normal distribution (Kolmogorov-Smirnov) in the sample. Quite a few of the 16 measures at 3 time-points violate Mauchly's test of sphericity.
I'm interested primarily in whether the subjects' scores on each of the 16 measures change over time, and whether the change is influenced by the grouping factors of location or day-length, or by their age or gender.
I had planned to do a repeated measures general linear model, with location, day-length, and gender as factors and age as a covariate. I have quite a lot of missing data points though, where subjects missed a question or weren't present for one of the time-points, and SPSS deletes their data on a listwise basis. My total sample size is 77 and with listwise deletion of missing values it's cut down to about 20! So I was then looking at doing a mixed model to cope with the missing data, with (I think) location, day-length, and time as fixed effects, and gender and age as random effects. But I'm not 100% confident this is appropriate.
So yeah, any advice would be gratefully received!