what will be the explanation of the moderation effect,
when the direct effect of an independent variable on the dependent variable is significant and positive.
while the moderated effect is also positive and significant, but the Beta of direct effect is high than the moderated effect.
You may also consider the P-Values and R² of both the direct and the moderated effect.
What is important is the direct effect of interaction that is significant, also see if the moderate effect is met in the low, medium and high level coefficients.
Please see the discussion of interaction (moderation) here. https://www.google.com/search?q=Montgomery+douglas+c+Design+and+Analysis+of+Experiments&oq=Montgomery+douglas+c+Design+and+Analysis+of+Experiments&aqs=chrome..69i57.36883j0j1&sourceid=chrome&ie=UTF-8
An interaction plot should answer your question. Best wishes, David Booth
Hi,
I very much agree that plotting the model is important for understanding the results. I find it difficult without doing so, as the combination of the predictor variables have to be taken into consideration simultaneously when interpret the findings. Substantively, interaction models is to study if a variable is related to the outcome differently at different levels of a third variable. For example, if a main group effect without a statistical significant interaction between the group and for example age, the group difference is present at all levels of age. The focus is now of course on the main effects. However, if the group x age interaction is statistical significant, the magnitude of the group difference is different over different levels of age. Now, the focus is not on the main effects, but the total model. In Hair and coauthors (Hair, J. F., Anderson, R. E., Tatham, R. L., & Black, W. C. (1998). Multivariate data analysis (5 ed.). New York: Prentice Hall), it is stated that the main effect in an interaction model with stat.sign interaction effect is a main effect of one variable when the other predictor variable is zero, while in a main effect model without interaction effect it is an effect of one variable (group) over all levels in the other predictor variable (age). This is the substantively reason for centering a continuous predictor variable around the mean in the interaction model, as the group difference now becomes when the sample is of mean age instead of being zero years old. Of course an age variable could be set to different age levels as well in order to explore how group differences are stat.sign. or not. Centering does also reduce multicollinearity. In sum, it is the combinations of all effects that are important here. And again, the plot will show this.
best,
Rolf
....... when both effects are positive, the higher levels of x1 is associated with higher levels of Y, with stronger relationship at higher than lower levels of x2.
Rolf
- Badges
- Science topic