How do I analyze the relative importance of moderator in one regression ? Any statistical limitations?
e.g. W and Z are both the moderator of the X-Y relationship. Both W and Z are continuous variables.
1. Could I know W is relative importance(stronger) than Z on positively moderating X-Y relationship?
2. If yes, how to conduct in spss or other software? Is it still plausible when one of moderators held no moderating effect on X-Y effect?
Could you please write these ideas in terms of your regression models so I at least would be clear on what you are asking. Best wishes David Booth
This is an interesting question. Perhaps you could compare the increase / change in R squared when adding one moderator versus the other relative to a model without interaction terms? I'm not sure whether there is a formal test of significance for the difference in two R squared change values but at least you could look to see which R squared change value is larger on a descriptive level.
It's an interesting question.
You can compare the ∆ beta coefficient and significant level of the interaction variable (i.e: Z) and the direct effect od independent variable (X) on dependent variable (Y). I hope this clue may help you to solve the problem. Cmiiw.
I think there is not just a statistical aspect to this question. How do you define "importance"? For example, imagine the effects of W and Z would be identical, as well as the effects of W*X and Z*X. Clearly, no statistical decision could be made, since all (conditional) effects would be the same. But if W and Z are interventions for example, W could be much easier and cheaper to implement than Z, therefore, ceteris paribus, W would be more important than Z.
Similar with your question, I think you should intepret how much W and Z change the relation of X-Y, how much is it possible to change W and Z itself and hence Y (for example by calculating how much change in Y would be possible at all, if you would take extreme values of W and Z, to get an impression) and then think about W and Z itself, how important these variables are for your research. Therefore, I think you cannot answer this question only in statistical terms.
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Sampling Design
- Sample Size