"Fighiting" with MIXED ANOVA!

[felice sirico]

Dear all,
I have a problem with the analysis of same complex (at least for me!) data. I try to summarize the scenario:

n=40
random assignment in 2 groups (treatment vs sham) (21 vs 19) (independent variable= GROUP coded 1 and 2).
outcome variable: time in milliseconds (SCORE).

each participant has been assessed two times (pre and post) (my whitin-subjects......correct?) (ASSESSMENT coded 1 and 2)
each participant in each group has been assessed for upper limb and lower imb in pre and post assessment (LIMB coded 1 and 2)
each participant in each group has been assessed for two types of stimulation: auditory and visual in pre and post assessment (STIMULATION coded 1 and 2)

So, I have a dataset like this:
ID=1 GROUP=1(treatment) ASSESSMENT=1(pre) LIMB=1(upper), STIMULATION=1(auditory) SCORE=millisecond
ID=1 GROUP=1(treatment) ASSESSMENT=2(post) LIMB=1(upper), STIMULATION=1(auditory) SCORE=millisecond
ID=1 GROUP=1(treatment) ASSESSMENT=1(pre) LIMB=1(upper), STIMULATION=2(visual) SCORE=millisecond
ID=1 GROUP=1(treatment) ASSESSMENT=2(post) LIMB=1(upper), STIMULATION=2(visual) SCORE=millisecond
ID=1 GROUP=1(treatment) ASSESSMENT=1(pre) LIMB=2(lower), STIMULATION=1(auditory) SCORE=millisecond
ID=1 GROUP=1(treatment) ASSESSMENT=2(post) LIMB=2(lower), STIMULATION=1(auditory) SCORE=millisecond
ID=1 GROUP=1(treatment) ASSESSMENT=1(pre) LIMB=2(lower), STIMULATION=2(visual) SCORE=millisecond
ID=1 GROUP=1(treatment) ASSESSMENT=2(post) LIMB=2(lower), STIMULATION=2(visual) SCORE=millisecond
And so on.....

I am not able to costruct the right command for STATA.
In my opinion I have a whitin-subjects factor (SCORE) as repeated measure in 2 time points AND GROUP, LIMB, STIMULATION as between-subjects factors.

Have these between.subjects factors to be considered has nested?
How can I write the right command in stata with error terms, nested variables and interactions???

I am able to write only "anova score".........after I have no idea because it is too much complex for me....
Thank you for your help

Felice

[Dave Airey]

It seems like you have a one between subjects factor and *three* within subjects factor design. There is some help here (https://www.stata.com/support/faqs/s...ova/#expand911) for a two within subjects design. You might look at each limb site separately to start with using that help at least.

[Bruce Weaver]

Hello Felice. Given that you have an experimental design with random assignment to groups, I suggest you consider using the post-treatment scores as the outcomes and the pre-treatment scores as covariates. See this short BMJ Stats Note, for example.

• https://www.bmj.com/content/323/7321/1123

HTH.

[felice sirico]

Dear Dave,
thank you so much for your post.......continuing with the reasoning I realized I misinterpreted the within and between subjects factor! Likely, your post confirms my errors and solves all doubts. I have to consider only the GROUP assigment as between subject factor. ASSESSMENT, LIMB, STIMULATION have to be considered as within subject factors!!! Now, I study your link and I try to perform to separate Mixed Anova (one for upper and one for lower limb).

Dear Bruce,
thank you so much for your post. Your proposed solution with ANCOVA approach is a perfect solution! Neverthless, I am trying to better understand the potential use of mixed ANOVA "putting all togheter"......

Thank you so much for your support!
Felice