I have a logistic regression model with one categorical predictor (four levels). The model does not fit well; I've used a Pearson's Chi Square Test and McFaddens Pseudo R squared to evaluate. I am comparing this model to two other models to compare which fits my data the best and would like to try to improve this model before I completely discount it. I've tried researching ways to improve model fit but I have not found how to transform a logistic model with a categorical predictor. Can this be done and how would I go about doing this?