Interpreting Interactions in Binary Logistic Regression

#1
Hi!
I was wondering if anyone knows how to interpret interactions in SPSS, for logistic regressions, or how a good source that might outline this. The interaction is between country level of development (0/1) and years in education (1-26). The interaction effect is significant, but the main effects are not. O.R is .923. Thank you very much!!
 

Mean Joe

TS Contributor
#2
I would say that there is an effect from years in education within level of development. Simply having a higher level of development is not enough; simply having more years in education is not enough. But having more years in education within one level of development is significant.

I can only point you to an online source: Plots distinguishing between main and interaction effects

I think you would be most interested in situation (e): note that some who exercise in the morning get more sleep, while others who exercise in the evening get more sleep (so no time of day effect). From the graph, some people on line 1 have more sleep than people on line 2, and vice versa. Also, there is no intensity of exercise effect--some with light intensity get more sleep, while others with heavy intensity get more sleep. From the graph, some people on the left side (light intensity) get more sleep than people on the right side, and vice versa.

An interaction effect is evidenced when the difference between the slopes changes significantly.
 
Last edited: