LOGISTIC REGRESSION ASSUMPTIONS - linearity for categorical IV's


I am going through the assumption for a logistic regression but I am stuck on making sure the 'IV is linearly related to the log odds'.

My DV is dichotomous and all of my IV's are categorical (some binary, others with multiple groups). I am struggling to test this assumption on my data as I do not know whether it must only be done on continuous IV's or it must still be done on categorical data. If it must be done on categorical IV's would anyone be able to advise how I can do this? (I'm using SPSS however, I'm new to this.. so apologies if this appears to be a silly question).

Many thanks!!


Less is more. Stay pure. Stay poor.

Linearity between a continuous variable and the logit is an assumption, but to my knowledge this assumption is implied to be true with categorical variable. In particular, think about a binary IV, you just draw a line between the to groups to get the coefficient. So that line will be linear and doesn't have the ability to be say, curvilinear since it is either one group or the other, there isn't a gradient of values it can take on.


Fortran must die
Even for OLS predictor categorical variables are always assumed to be linear. This is the nature of categorical variables.