Hi I am doing mlogit and forward selection at the moment. I have categorical and continuous variables. Categorical variables are each coded with more than two dummies. There is a problem when you add categorical variables into the model you can't tell from the significance of the individual dummies that will the entire categorical variable be significant or not. So is there some method to compare two models to each other and find out is the model better with or without the dummies? If there is, could you help me with code...? Ville [[alternative HTML version deleted]]