How to perform ANCOVA

#1
Hello All,

my study is about the effect of stress on donation behavior.

Using TSST for stress induction (+control task), a donation task with 80 charities (40 charities in the pre- and 40 charities in the post stress conditon) in which the Participants had the opportunity to donate between 0 and 20 Euro. Additionally Empathy, Compassion and Perspective taking was measured in the pre- and post stress condition. 43 Participants. Duration of 174 minutes.
Cortisol, pulse, and blood pressure was measured over the 5 time points as a measurement of stress reaction.

I want to find out if the intensity of the expression of empathy, mentalizing, and perspective taking(IV) has an moderating effect on donation sums(DV).
I would like to perform an ANCOVA with the post stress condition as the dependent variables. Do you think it is right?


I cant figure out how to place all these variables.

Stress and Control group (I am not sure if I should use this as a Fixed Factor)
Mean donation (pre-stress) (Covariate)
Mean donation (post-stress) (DV)

Mean Empathy (pre-stress)
Empathy-post (post-stress)
Perspective taking (pre-stress)
Perspective taking (post-stress)
Compassion (pre-stress)
Compassion (post-stress)
Should I use the pre- or the post-stress outcomes of empathy, compassion and perspective taking as further covariates?
 

Karabiner

TS Contributor
#2
Why do you think you have to put all variables into 1 model?

As a start, you could perform a series of "mixed" repeated-measures analyses of variance,
with group as between-subjects factor, time of measurement (pre/post) as repreated-
measures factor, and with mean donation, empathy, perspective taking, and compassion
as the dependent variables, respectively. The interaction between group and time
will tell you whether e.g. changes amount of donations between over time were different
between conditions.

How then to combine donation, empathy etc. in just one model depends on what
precisely you want to find out, your theoretical assumptions.

With kind regards

Karabiner
 
#3
Guten Tag Karabiner,

thanks again for your help!
empathy, compassion etc. do not have to be in one model. I can perform separate models and yet I want to know if they have a moderating effect on the donated amouts. That's why I am insecure which variable (empathy pre- or post-TSST) to use and where to put it..
 

Karabiner

TS Contributor
#5
I suppose that compassion etc. are mediators, not moderators (if they are measurements of psychological states). I am not sure how that can be modeled. Maybe by using multilevel modeling where donation is regressed on empathy etc., or maybe by using a macro such as MEMORE as mentioned here: http://afhayes.com/spss-sas-and-r-macros-and-code.html

But if instead they are trait variables, then they could be moderators, and I'd suppose that the baseline value is of promary interest.
But then I would ask why a state variable was measured twice.

Regarding homogenity of variances, as far as I know it does not hurt too much if groups are if equal size.

With kind regards

Karabiner
 
#6
Thank you a thousand times!
You are right they are mediators and I have also state variables for empathy etc. Just found them because to your question.
I will try to use multileven modeling. Thank you very much!