Hi everyone,
I am conducting research for my Master's thesis. I am using PROCESS by Johnson-Neyman to analyze my Moderator model. I test the relationship between Public Service Motivation and Public Sector Preference, moderator variable: Love of Money (LOM).
However, when I run the analysis, it gives the result below with two zones of significance. I really do not understand what's wrong and what should I interpret?
I really really need your help.
Thank you very much.
There is nothing wrong with it. Important is the effect size under "Effect", it represents the conditional effect of your predictor on the DV, i.e. dependent on values of your moderator. It does not matter if it is positive or negative. It is not exactly symmetric, since the SE are slightly different.
The regression coefficient for the interaction shows how much the effect of your predictor changes with changing values of your moderator by one unit (not depicted). Here you see a whole range of values for your moderator from -2.9034 to +.296, so the conditional effect of your predictor on the DV changes from -.5169 to +.49. Therefore it may be concluded that the regression weight for the interaction is around 0.3147 --> how much the effect changes with every unit change of the moderator. Apparently the conditional effect changes and switches sign. Cosequentely, effects around zero are not significant.
Quynh Vu, please note that the PROCESS macro is the work of Andrew Hayes, not Johnson-Neyman.
- https://www.afhayes.com/index.html
Second, it is a bit wrong-headed, IMO, to look at the Johnson-Neyman output to find the smallest conditional effects that are statistically significant. I think it is often (if not always) more sensible to use content expertise to determine a priori the smallest conditional effects of interest, and then check whether conditional effects of that magnitude are statistically significant. YMMV.
- Badges
- Science topic
- Similar topics
- Biological Science
- Kinesiology
- Running