Hello,
I am looking for help for analyzing my experimental data.
Here is my problem:
I conducted an experiment where participants were randomly assigned to three different groups. Thr first group (X1) received a positive stimulus, the second group (X2) received a negative stimulus and the third group (X0 = control group) did not receive any stimulus at all.
The dependent variable (Y1) is an attitudinal construct measured by 10 indicators. My assumptions are that the positive stimulus has a positive direct effect on Y1 and the negative stimulus has a negative direct effect on Y1.
Yet, there is the problem that there is a second dependent variable (Y2) which is also affected by the treatment factor. Y2 is measured by 5 indicators. Aditionally, Y2 is assumed to influence Y1. Thus, I suppose that this constellation could be well-suited for mediation analysis. Or to put it another way: I assume that the treatment factor not only has direct effects on Y1 but also an indirect effect via the mediator Y2.
Because the treatment variable is a categorial variable with three groups, I thought I could generate two dummy-variables for X1 (1=positive stimulus; 0=no positive stimulus) and X2 (1=negative stimulus; 0=no negative stimulus) and integrate them both into a structural model. As I understood, it is not possible to generate three dummy-variables for every treatment status because of collinearity?!
Now, I want to test the following hypotheses within one model:
X1 -> Y1 (+)
X2 -> Y1 (-)
X1 -> Y2 (+)
X2 -> Y2 (-)
Y2 -> Y1 (+)
In addition to these direct effects, I want to test the total effects of X1 and X2 on Y1.
Since I am not familiar with analyzing such constellations in SmartPLS, I would like to ask if anybody can tell me if my approach with two dummy-variables is reasonable or complete nonsense.
Thank you very much for your help!
Best regards,
Chris