To figure out the strength of effect for the variable in a regression equation, which one is more appropriate to report for the predictors: Beta or semi-partial correlation? And, in the case of using dummy coded variables, will the answer be the same?
Good question. Researchers mostly report the beta coefficient because they can make the statement that "this is the effect of X on Y holding all other covariates constant". There is some debate as to whether this statement is correct, but nonetheless, this is the language that is most often used. Reporting partial correlations may be a discipline specific thing, but in health care research (my discipline) this is rarely (if ever) reported. When using dummy variables then the coefficient for that dummy will provide the effect (or relationship, depending on context) of X on Y, after controlling for the other covariates in the model.
Thank you Dr. Linden for your answer. I agree with using Beta rather than the semipartial correlation in reporting the variable effect in MLR
- Badges
- Science topic
- Similar topics
- Mathematics
- Statistics
- Regression