Accounting for Covariates: which is the best test and how to do this in matlab

Hello

I would really appreciate some help. I have been trying to find out which is the best test to account for Covariates. I have a repeated measures design. For instance participants will see on of two stimuli, and we will collect two or more continuous data relating to that stimuli. For instance how long they perceived the stimuli to last and the time it took to respond. There are a few other similar continous stimuli.

I have done repeated measures t test for all of these dependant variables. However, I am mostly interested how long they perceived the stimuli to last. I now want to see if my data is still significant considering the other continous data as possible covariates. (FYI this is just an example)

From reading around (although it seems unclear) I think I need a mancova. However can you advise what is the best test to run and how to run it in matlab (or point me in the direction)?

Answer 1

[Adam Danz]

You could perform an analysis of covariance using anovan() and explicitly identiying the covariates. Here's an example from data I recently analyzed,

[p,tbl,stats,terms] = anovan(y,{cc, factor}, 'Continuous',1,'model','interaction',...
    'varnames',{'VarName1','VarName2'})

where cc is the continuous covariate and factor is the categorical variable. Importantly, note that the Continuous property indicates that the first variable in the 2nd input is continuous. If there's more than 1 continuous covariate, use a vector of indices. This is all described in the documentation, as well.

[Rayvara123]

Hello, thanks for this. Would this work for a repeated measures design?

[Adam Danz]

Could you explain your design in more detail? It's not clear to me what the repeated measure is and what the covariate is. Specifically, does the covariate vary between the two repeated measures?

http://imaging.mrc-cbu.cam.ac.uk/statswiki/FAQ/constantcov

[Rayvara123]

Ah yes. So imagine my participant is reading about two people. One person has depression and one person has anxiety. For each person they need to seperatly decide, if they would like to give them a certain amount of money (between 1 and 10). And then also rate them on a few factors like how trustworthy they are and how friendly they are.

The iv is the type of my condition. Dv is the money they give each person. And the covariate is how friendly and trustworthy they are.

Does that make sense? Let me know if you need more info.

[Adam Danz]

That makes sense but what is the repeated measure?

[Rayvara123]

All participants will rate all the case studies. They will rate the person with depression and the person with anxiety, and any other mh illness.

Is this not something you would consider repeated measures?

[Rayvara123]

This is as opposed to option where some participants rate the case with anxiety, and the rest of the participants rating the case with depression

[Adam Danz]

That is a repeated measures and covariates (money and ratings) could potentially vary between the mh groups so the link I shared on constant covariates does not apply.

From your question, ' I am mostly interested how long they perceived the stimuli to last. ', but I don't see that variable in your description.

What's the exact question you're asking because that will determine which statistic to use.