My research question has its independent variable as WC and dependent variable as AQ. Now I am adding a interaction by introducing my moderator BF. The interaction term is WC*BF. However, I did a chi square test for WC and BF to see if there are independent, and the results showed they are signidicantly correlated with each other.
What should I do in this situation?
Hello Yani,
The usual concerns in such settings surround: (a) confounding, especially in studies in which no direct manipulation of variables has taken place; and (b) biased estimators of regression coefficients (and artificially large standard errors). So-called instrumental variables might be helpful, but may not be identifiable or accessible for a given scenario. Larger N can help with the upward biased SE values, of course.
From a practical perspective:
Other than nonsensical textbook examples (e.g., foot size and intellectual functioning), it's difficult to find variables in real life that are completely uncorrelated. So, you have an IV and a moderator which are correlated. That does not prevent you from investigating whether the moderator value affects the IV-DV relationship in a meaningful way. Do be aware that, if confounding (with another or other variables) is the reason for the IV-M relationship, you should avoid, or at the very least, qualify any inferences about "cause."
Good luck with your work.
if BF is a qualitative variable then it can be used as a factor to do Grouped regression otherwise use expand WC*BF to "WC+WC.BF""
>>>you can look at AQ vs WC using quantile regression for slope change over your BF range
- Badges
- Science topic