I am writing a short paper on voter turnout. I would like to use bianary logistic regression for this purpouse. I have one problem, my household income variable is coded in 16 categories (5000€ each). Because of a relatively small sample, I can't afford to use it as a categorical variable, because it would mean 15 predictors more in the model and I think that there is some rule of thumb that I need at least 10 cases in each category of the dependent variable for each predictor to prevent overfiting the model)
Because of this reasin, I would like to use this categorised variable as a continuous one (so that I don't have to use dummy variables), is that legit?
Can I treat this variable as a cardinal one, if there are so many categories?
Advertise on Talk Stats