0

I am conducting a correlation study and have developed a logit model. I would like some help interpreting the interaction variable's coefficient.

The coefficient of the interaction term smoking*female, where female =1, is 0.16

How can I interpret this?

nwaldo
  • 531
  • 2
  • 11
Sabine
  • 35
  • 5
  • 1
    Take a look at this site: https://stats.idre.ucla.edu/stata/seminars/deciphering-interactions-in-logistic-regression/ – ColorStatistics Apr 12 '20 at 11:42
  • Although not written in the context of logistic regression, it might help you to read: [Interpretation of betas when there are multiple categorical variables](https://stats.stackexchange.com/a/120035/7290), & [Interpretation of interaction term](https://stats.stackexchange.com/a/122251/7290). – gung - Reinstate Monica Apr 12 '20 at 21:37

1 Answers1

1

The coefficient on an interaction represents the difference between the effect of one variable at the reference level of the other and the effect of that variable at the non-reference level of the other. In your case, it represents the difference in the effect of smoke between males and females. In particular, because the coefficient is negative and the reference group is males, this means that the effect of smoking on the log odds of the event is 0,16 smaller for females than it is for males. Put another way, the difference in the log odds of the event between smokers and nonsmokers is 0,16 lower for females than it is for males.

Noah
  • 20,638
  • 2
  • 20
  • 58
  • Thank you very much, this helps a lot. My only question is since it is a correlational study can i use this same interpretation? – Sabine Apr 13 '20 at 09:54
  • Focus on "differences" rather than "effects" since "effects" often evokes a causal interpretation. The last sentence of my answer doesn't make any causal claims, it just describes the observed pattern of results in the data. – Noah Apr 13 '20 at 17:28