Analysis for a repeated measures design with covariate that varies over conditions?

[KVella]

Hi all,

I'm trying to find the correct analysis that will allow me to carry out the equivalent of a repeated measures MANCOVA with a covariate that varies over three conditions. Basically I have an experiment with 3 different types of task where every participant does every task in a randomised order. I'm measuring the effect of these tasks on a bunch of mood measures. BUT I've found out that one of the tasks was significantly more difficult than the other two. I think this means I need to make task difficulty a covariate, however it is not constant over the three conditions.

Does anyone know what kind of analysis I should be running, and how I should be setting my data up (wide or long form). Is it some kind of mixed modelling (of which I know very little)? Bonus points if it can be done using SPSS.

Much thanks to anyone who answers,
Kellie

 

[dhvalden]

Re: Analysis for a repeated measures design with covariate that varies over condition

hi! i'm new around here but a see that many people see your post and no one reply, just like mine, the pots below yours, so lets do something about it.

First, What do mean that the difficulty of the task is not constant over the conditions?? for what I understand that make it a variable that can be used in any way you want, or i'm I wrong?

Anyway, I think you gonna need to move from a mancova framework to a multiple regression with interaction, in R (lavaan) o Mplus environment. In Mplus forums there are various peoples asking how to conduct analysis like yours. Maybe you can check there!

 

[KVella]

Re: Analysis for a repeated measures design with covariate that varies over condition

Thanks for responding!

Just to clarify:

I gave my participants 3 tasks to do (same task, but one was alone, one was with another participant, and one with a computer-generated character). It turned out that doing the task alone was more difficult than it was in the other two conditions. However difficulty is not the DV of interest, mood is. Understandably though, difficulty could influence mood so I need to take it into account in whatever analysis I do. I measured difficulty by getting the participants to rate the experience on a 3-item, 7-point likert scale after every task. Mood is also measured on a likert scale.

My research question is 'does mood differ depending on the how the task is carried out?'

I'm not sure how regression would help me (but I'm all ears), as I'm not wanting to know if task predicted mood but whether the tasks differed in level of mood generated. Does that make sense? I'll have a look at your question but I'm not sure how much I can help as I'm a newbie in these parts.

 

[dhvalden]

Re: Analysis for a repeated measures design with covariate that varies over condition

Well, this is what I would do if I was you, so just take it as an option.

Well do you have the date for al the participanst in just one big data frame, isn't? And I imagine you have a variable indicating to which group the participant belongs. So, you could consider the grouping variable as another predictor in a multiple regression, by transforming the grouping variable in 3 binary or dummy variables, each one predicting if being member of some group predict the mood of the participants with individual betas for each membership. Also, now that your using a regression analysis you can control for every variable you want, including difficulty of the task.

But also now, you can use regression to determine if the difficulty can moderate o mediate the results in the model or the mood moderate of mediate the perception of difficulty.