MANOVA

Moderators: statman, Analyst Techy, andris, Fierce, GerineL, Smash

richiepap
Posts: 2
Joined: Tue Feb 02, 2016 7:23 pm

MANOVA

Postby richiepap » Tue Feb 02, 2016 7:49 pm

Hi everyone, hope I'm posting this in the right place.

I am conducting an RCT with one categorical IV which has three levels/conditions, and two scale DVs. Participants in each condition are randomised from the same population - only the intervention they use will vary. For each condition I will take pretest measurements of both DVs (which aren't expected to differ to any significant degree between conditions), run the experiment, and then take posttest measurements of both DVs.

Where I am getting confused is, I'm not 100% clear in my head whether to check for significant differences between the posttest scores, or significant differences between the difference (posttest-minus-pretest) scores.

I'm not sure if I've explained that very well! But I'm willing to have another try if I've not made myself clear.

Thanks in advance for any responses.

Rich.
GerineL
Moderator
Posts: 1477
Joined: Tue Jun 10, 2008 4:50 pm

Re: MANOVA

Postby GerineL » Tue Feb 09, 2016 3:41 pm

use repeated measures analysis. you can see differences between groups both at pre and post, the effect of time, and whether the changes over time are different between conditions.
GerineL
Moderator
Posts: 1477
Joined: Tue Jun 10, 2008 4:50 pm

Re: MANOVA

Postby GerineL » Tue Feb 09, 2016 3:42 pm

I really strongly recommend doing this by the way, and I also strongly recommend checking whether or not they differ at pre test, it happens very often that you randomly select participants but that they do differ at pretest.
richiepap
Posts: 2
Joined: Tue Feb 02, 2016 7:23 pm

Re: MANOVA

Postby richiepap » Sun Mar 06, 2016 10:41 am

Hey GerineL, thank you for your replies. I did a quick check to see if there were any significant differences at pretest, and there were! I was not expecting that, but it's good to see you say that it's a common thing. Anyway, after a bit of tweaking I've got a pretty ok dataset for parametric analyses, although it's rather small at 32 participants.

Just out of interest, does anyone know if a repeated measures MANCOVA is a thing? Because this is my basic design, and honestly I'm so confused about what to do with it!

Image

The pretest and posttest sample sizes are the same, and the between subjects sample sizes are very similar. At this point I can't be sure if the follow up will end up with the same sizes, or even remotely close, as I am still collecting data - but it doesn't really matter if the follow up doesn't get included, it was just a nice addition.

Analysing pre- to posttest Believability and Discomfort change (which are posited as closely interrelated variables, hence the MANOVA part), and the comparison between three groups to see which peformed better/worse, are the key requirements for my study. As I say, the followup isn't too important. I would also like to explore what role CFQ has, as it is posited as a measure of the capacity to be able to change Believability and Discomfort, thus thought it might be included as a covariate. I've found no good reason yet to include the measures of STAI, BDI-II, and Word Syllables, and so they can effectively be disregarded.

I honestly love stats, and would love to be great at them, but they do confuse the hell out of me :? :)
GerineL
Moderator
Posts: 1477
Joined: Tue Jun 10, 2008 4:50 pm

Re: MANOVA

Postby GerineL » Mon Mar 07, 2016 12:02 pm

Yes, you can add multiple outcome measures in Repeated measures. Yet, I do think the model you are trying to run may be way too heavy for the number of cases you have, you almost certainly have a power problem. Therefore, you may want to run them separately after all.

Who is online

Users browsing this forum: No registered users and 1 guest

cron