0
$\begingroup$

I plan to investigate the effect of a personality trait on reaction measures (emotional reactions and intention for political participation) in a vignette study with two conditions. I assume that the condition moderates the relationship between personality trait and reaction only as long as potential confounding variables such as political ideology and political interest are not controlled for.

Which statistical method is best suited to test this hypothesis? A moderated regression analysis? How would I include control variables in such a moderated regression model?

Improve this question
$\endgroup$

1 Answer 1

Reset to default
1
$\begingroup$

You can include them by ... well, by including them. Let's call the trait $X_1$ and the moderator $M$ and the confounders $X_2 \dots X_P$. Then:

Moderated regression without confounders:

$Y~X_1 + M + X_1*M$

Add confounders:

$Y~X_1 + M + X_1M + X_2 + X_3 + \dots X_p$

Then you can examine all the coefficients and test them, as well.

This assumes you meant moderation, which is another name for interaction. If you really meant mediation, things are different. The two terms (moderation and mediation) are often confused with each other. Not that I am saying you are doing this.

Improve this answer
$\endgroup$
4
  • $\begingroup$ Thank you for the explanation! Indeed, I meant moderation. What I ask myself though if it is the best way to include confounders in the regression model. Wouldn't that cause multicolinearity? So basically my question is: Is it the smartest way to use moderated regression for this type of hypothesis (interaction only exists if not controlled for possible confounds). Or is it smarter to use a different statistical approach? $\endgroup$
    – al01
    Commented Feb 29 at 18:26
  • $\begingroup$ Confounders may or may not cause collinearity. That should be assessed using condition indexes or VIF. I think your method is basically fine. $\endgroup$
    – Peter Flom
    Commented Feb 29 at 18:29
  • $\begingroup$ Thank you very much for your help :) $\endgroup$
    – al01
    Commented Mar 2 at 8:41
  • $\begingroup$ You're welcome. If my answer resolves the issue for you, you can accept it by clicking the check mark. $\endgroup$
    – Peter Flom
    Commented Mar 2 at 12:25

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.