The continuous data are log-transformed for logistic regression.
But how to report the OR of the interaction term "Zn x Mn" for example?
For the individual variable Mn e.g. I would mention "the OR of Mn on a logarithmic scale is ... per mcg/l" or "for every change of factor 10 increase in Zn concentration, the OR is … per mcg/l" (which one is best by the way?), but for an interaction, how do you report this?
Thx!
Daan, if you are asking how to interpret the coefficients or the Exp(B) values, perhaps the attached excerpts from Jaccard (2001) will help.
Jaccard, J. (2001). Interaction effects in logistic regression. Series: Quantitative applications in the Social Sciences.
Yes, thanks a lot for this!
I'm more specifically trying to figure out how to mention it in a research article.
It sounds strange to say "The interaction term Mn x Zn has an odds ratio of ... given a 1 unit increase in Mn when Zn = 0", since both are metal values in a biological matrix and none of them will ever really be zero of course?
How would I say that then?
You could always center Mn and Zn on possible values before estimating your model. If you do that, you would say that "the interaction term Mn x Zn has an odds ratio of ... given a 1 unit increase in Mn when Zn = K", where K is the value you centered on.
- Badges
- Science topic
- Similar topics
- Linguistics
- Composition Studies
- Publication