I've been trying to find specific guidelines for running a repeated measures ANCOVA in SPSS. The important thing is that I don't have any between-subjects variables, only one within-subjects variable and one covariate.
So in short, my design is:
- dependent variable: reaction time
- 1 within-subjects condition with 2 levels: before and after treatment
- 1 covariate: subject's BMI
Is it possible to just use the option "Analyse>General linear models>Repeated measurres..." and leave the between-subjects factor field empty?
If yes, (how) do I have to check the "parallel regression slopes" assumption?
Also, when I use the option above, the output I get is quit confusing to me. In what exact tables can I check:
- The significance and effect size of the within condition?
- The significance and effect size of the covariate?
- How the within-condition and covariate interact?
Really hope someone can help me with this!
Run the regression: After=a*before +b*BMI +Eps this is ANCOVA in Repeated measures form. For a ref see https://www.bing.com/search?q=montgomery+design+and+analysis+of+experiments&qs=SC&pq=montgomer+design+and&sk=SC2&sc=3-20&cvid=7BE9CEC687824B79B1FE14A83B02A19F&FORM=QBLH&sp=3&ghc=1
BTW forget SPSS. SEE this link for a much easier and sensible package than SPSS-look at multiple regression: http://sgpwe.izt.uam.mx/files/users/uami/gma/R_for_dummies.pdf
David Booth
I do not understand why you would want to use an ANCOVA in your case, in which you do not have a between subjects group factor. An ANCOVA is useful when you want to control for possible differences between groups on the pre-test measure.
See:
https://www.theanalysisfactor.com/ancova-assumptions-when-slopes-are-unequal/
In a repeated measures design a between-subject covariate is usually uninteresting because it just explains variance in the within-subjects error term. It gets a little more interesting if it interacts with other predictors, so it can be useful to fit that model. This is also conventionally the test of parallel slopes ... (though I find that way of thinking needlessly confusing - its just the inclusion of an interaction term).
In SPSS I'd reccommend centering the covariate manually and computing interaction terms with the covariate manually before entering them in the model. This should greatly aid interpretation.
If you have a time-varying covariate then I'd use a multilevel (aka linear mixed model) rather than ANOVA.
Hello Eva van Dueren den Hollander, I have the same design as you and I'm stuck with the same issue as you. May I ask what did you end up doing? I'm also very confused especially with checking the assumptions.
- Badges
- Science topic
- Similar topics
- Social Psychology
- Attitude