I am running mediation analysis using the PROCESS macro for SPSS and I have binary independent variables. It is my understanding that Haynes (2017) recommends that you report unstandardized beta coefficients. Can anyone help my explain why we would report these instead of standardized?
For a binary independent (X) variable, it makes little sense to say "for a one-standard deviation unit change in X, we expect an xx standard deviation change in Y." Instead, we interpret the unstandardized coefficient, which gives the mean difference between the two categories/groups on the dependent variable Y.
Thank you so much , a little clarification did you mean to say which gives the mean difference between the two categories/groups on x? Or was it y?
Christian Geiser I have a question regarding the interpretation of mediation results when using a binary independent variable. Let's say I am looking at the mediating effect of parental concerns (M) on the relationship between parental visible minority status (X) and children's physical activity (Y). Parental visible minority status is a binary variable (coded 1 for visible minority and 2 for nonvisible minority). In my mediation analysis, I get -0.25 for my path A ( X's effect on M). How would I interpret this value? I understand that -0.25 is a difference of means on the mediator but how would I interpret this in a written format and verbally?
The path coefficient of -0.25 between the binary X variable (minority status) and the continuous M variable (parental concerns) indicates that the two minority status groups differ with regard to their average parental concerns with the sample mean difference being equal to -0.25. Specifically, the nonvisible minority group (which is coded with the higher value = 2) has a smaller mean level of parental concern than the visible minority group (i.e., the group that is coded with the smaller value = 1). You can see that from the negative sign of the path coefficient (-0.25).
You can verify this result by computing the means for the two minority groups. The visible minority group should have a higher sample mean for parental concern than the nonvisible minority group, and when you take the difference between the sample means you should get -0.25.
- Badges
- Science topic
- Similar topics
- Statistics
- Survey