Type of predictors for logistic regression Can predictors in logistic regression be categorical, numerical and ordinal?
If categorical, can they be trichotomous?
 A: The situation is the same as for ordinary linear regression. Namely, the model allows you to posit a linear relationship between an IV and the (log odds of the) DV, which is appropriate for many numerical predictors, and also suffices for categorical predictors on a nominal scale if you code each predictor into k - 1 dummy variables where k is the number of levels (don't forget to include an intercept term). The ordinal case is more difficult, because inserting an ordinal predictor as a single model term treats the differences between levels as equal, and dummy-coding an ordinal predictor makes the order information unavailable to the model. If you need more sophistication than either of those methods, you'll need a fancier model. Here's an article that discusses the question of what to do with one kind of ordinal predictor, rating scales:
Tutz, G., & Gertheiss, J. (2014). Rating scales as predictors—The old question of scale level and some answers. Psychometrika, 79, 357–376. doi:10.1007/s11336-013-9343-3
