I'm running some basic regression models where some of my models have continuous DVs and others have categorical DVs.
I've noticed that for my categorical DV models, my BIC values are generally positive and large (e.g., 10469.264), whereas for my continuous DV models, my BIC values are negative (e.g., -672.959).
My questions: 1) Is this a coincidence, or are BIC values for continuous DV models always negative while BIC values for categorical DV models are always positive? If not a coincidence, why is this the case?
2) When comparing BIC values between models, with models with positive BIC values, I am under the assumption that the model with the smaller (i.e., closer to "0") BIC value is the better fitting model. With models with negative BIC values, is the model with the more negative BIC value the better fitting model (e.g., -10 fits better than -5), or is the model with the BIC value closer to "0" the better fitting model? I'm assuming the former, but I wanted to be sure.
Other details of my models: estimator = MLR, TYPE is complex. IVs are factor scores.