I have this example logit model where some of the variables are factors but I'm not too sure how to interpret the effects. If I understand logit models correctly the coefficients that we get from the fitted model are the change in log-odds per unit change in the explanatory variable holding everything else constant.

If take the exp() of the coefficients then I have the odds. I'm interested in the impact over the probability of some cathegorical variable to be "male" or "female" for instance. Could you please help me understand this? if allEffects() is not what I'm looking for, could you please let me know how could I get them?


titanic <- glm(survived ~ passengerClass + sex + age,data=Titanic,
titanic.all <- allEffects(titanic, typical=median, 
                          passengerClass3rd=1/3, sexmale=0.5))
     ticks=list(at=c(.01, .05, seq(.1, .9, by=.2), .95, .99)),

EDIT: I don't think it is a duplicate. I'm interested in the output of the package effects, in particular in the output of the function allEffects.

I found one document with the following notes "Notice that the print method for the object returned by allEffects reports tables of the effects, which, by default, are on the scale of the response variable, for a logit model, on the probability scale"

  • $\begingroup$ You will find this answered in many threads: search on logistic interpretation categorical. $\endgroup$
    – whuber
    Apr 18 '13 at 20:02
  • $\begingroup$ @whuber thanks for the link. I'm particularly interested in the effects package. What is exactly the interpretation of allEffects()function? $\endgroup$
    – nopeva
    Apr 18 '13 at 20:35
  • $\begingroup$ I think interpreting what a function is supposed to do would likely be off topic for CrossValidated (stats.SE), perhaps belonging on SO. Perhaps you could edit your question to be both more clearly within the range of topics suitable for CV and to more clearly distinguish it from previous questions. $\endgroup$
    – Glen_b
    Apr 18 '13 at 23:15

Browse other questions tagged or ask your own question.