I used PROCESS model 1 to calculate a moderation effect. The results show a significant interaction between the variables, but the conditional effects are non-significant. I mean standardized the moderator (Zmean_Tr).
How should I interpret these result? Could this be due to an error in my data? Or should I just conclude the interaction is not significant after all?
Many thanks!
Your data seems totally ok, but your interpretation is flawed. The significant interaction only tells you that the slopes of your X variable change with changes in your moderator W. Since your exploratory effect of the interaction is very small (2%) I would not expect strong changes.
Additionally, the test of the conditional effects at different levels of the moderator does not compare the slopes with each other, but if each slope is different from zero. Therefore, the slopes may differ from each other but all not from zero. And again additionally, the values for the moderator to test the slopes are artificial (unless you specified it by yourself), in PROCESS 16th, 50th and 84th percentile, others use +/- 1 SD and in case of not symmetrical distributions, this will again change the results. As you can see, with increasing values of W the conditional effect switched from negative to positive values (therefore crossing zero, why they are close to zero and not significant) and has the largest value at W = 1,073. Imagine you would test W = 2.0 the slope would surely be significant.
Taken together, it is not a problem of the data, but the interpretation.
Thanks a lot for your clear and helpful explanation to my analyses! :) Hopefully this discussion can be useful to others in the future as well.
- Badges
- Science topic
- Similar topics
- Mathematics