Mixed repeated measures ANOVA with covariate
Hi all,
I am new here and having lots of stress with these stats! I have done a mixed repeated measures ANOVA in JASP, with trial type (2 levels) as a within subjects factor, group (2 levels) as a between subjects factor, and age as a covariate.
I have no idea how to interpret this table! Or even if I have run the test correctly. I have not had to use covariates before and I have no other avenue to ask for support. I would really appreciate any help.
Thanks
Comments
Hi,
If I read your table you have only a main effect of GRP according to classical view of p-value (i.e., p-value < .05 lead to accept H1 and reject H0).
Upstairs, you can see the effect of within subject factors and interactions. Downstairs, you can see the effect of between subject factors.
If you need more help tell me
I'm still a student, so perhaps more competent people could provide more information.
Have a nice weekend
Kevin
To me, it looks like you put AGE in as a "between subjects factor," not a covariate. Can you confirm that you put AGE in the covariate box? Also, it would be helpful to know how AGE was measured. Is it actual "years old" or some reduced set of age ranges?
Hi,
Thanks for responding, I really appreciate it. I've attached an image of the whole screen so you can see where I put the variables. I initially did a mixed repeated measures ANOVA, trial type x group, but I've been asked to control for age, so I assumed it would just be a case of running the ANOVA again, adding age in the covariate section, so that's hopefully what I've done here. Perhaps it's not that simple? :-)
Age is a continuous variable, measured in years - I've included descriptives here too if that helps.
Thanks again for responding!
Hmm I am a little confused -- in the Descriptives table, Age gets a 0 or a 1...can you screenshot a few rows of the data spreadsheet?
Cheers,
E.J.
Hi,
It seems that in the descriptives table, Claarek specified the descriptives of Age, split by another variable that has the levels 0 and 1 (the table does not show which variable that is, maybe that is good to add to JASP so the table becomes more informative; i.e., a note that says "split by <variable>).
As for including a covariate: you can take a binary variable, tell JASP it's continuous, and add it as a covariate, or whether you tell JASP it's nominal, and add it as a between subjects factor. The default model terms do differ based on whether it's a covariate or between subjects factor, but if you manually make the model terms the same by adding the same interactions, you get identical output:
The data here are exactly the same, but in the second screenshot I took the binary variable contBinom, told JASP to treat it as continuous and added it as a covariate instead of a between subjects term.
The interpretation here is the same: do you other factors still explain some variance of the dependent variable, when you take into account the effect of other model terms? This why you would include these model terms (these could be other within subject factors, between subject factors, or continuous covariates) in your analysis. Adding these factors "controls" for their influence, in order to rule out confounding effects. You can read more about why we do this in my response here (see point 3): https://forum.cogsci.nl/discussion/5940/help-for-covariate-analysis-with-mixed-anova#latest
In your specific analyisis, it seems that there is minor support for a difference between the "GRP" levels (p < 0.05, which is really not that much evidence - we are on a Bayesian forum after all ;-)). However, for the other variables, it does not seem that there are group differences.
Does this clarify things? If not, please let me know and I can elaborate!
Cheers,
Johnny