Example: In a study on academic performance, the student's age (independent variable) may not have a direct impact on grades (intercept), but when interacting with the amount of study hours (moderator variable), age may significantly influence the final grades.
What may be the main cause of this case?
Yes, this can happen when the effect of the IV on the DV has a different sign for low versus high values of the moderator (e.g., a positive slope for low values of the moderator and a negative slope for high values of the moderator). Then, the effects may cancel out when the moderator is not included in the regression model.
The easiest way to understand what is going on is to plot the regression lines for the regression of Y on X for different values of the moderator. That way, you can see in which way the relationship between X and Y may differ across different values of the moderator.
Dear Dr. Geiser,
Thank you very much for your insightful answer, I suppose in this case a simple slope analysis should be performed including the test for statistical difference between the two slopes (low values & high values) of the moderator?.
Kind regards,
Muziwandile Luthuli.
You can test the effect of your moderator variable as an interaction effect with your IV. Because they are both interval-level variables, the interpretation can be a bit tricky. I recommend that you do a search on "linear by linear interaction."
- Badges
- Science topic
- Similar topics
- Biological Science
- Molecular Cell Biology
- Culture Cells
- Cell Line