I am designing a moderated mediation model (Hayes' Model 7). I want to use a 5-point Likert scale as a moderator variable in my model. Will this cause a problem? Or do I need to make an adjustment to use a continuous variable as a moderator? Because in some sources it is stated that the continuous variable should be made categorical. Thank you very much in advance for your answers.
Hello Erdi,
Moderation simply implies that there is an interactive effect of a target IV with a supposed moderating variable. Nothing about this requires that the moderating variable be discrete/categorical, or continuous in form.
However, ordinal scales require the most arithmetic, as you can't presume a strictly linear function as you can when a variable is either dichotomous or continuous.
For your situation, the test of the interaction hypothesis could be evaluated by inserting 4 dummy coded variables for the moderator, and the product of each of these with the IV as the actual information for the presence/absence of interaction. So the test for interaction would minimally be a 4 df effect (more if the IV was categorical and had more than two levels).
Good luck with your work.
Dear professor, thank you very much for your explanations. My indecision has become clearer. Yours sincerely,
Dear Erdi,
As Prof. Morse commented, "Nothing about this requires that the moderating variable be discrete/categorical, or continuous in form."
However, in most studies, 5-point Likert-type scale is treated as a continuous variable. You can test it as a continuous variable.
Dear Dr. Kim,
Thank you very much for your response. I wanted to get your opinion as I came across some confusing statements in different sources. Thank you again.
Sincerely yours,
Erdi
- Badges
- Science method
- Similar topics
- Mathematics
- Statistics
- Statistical Techniques
- Meta-Analysis