Hi,
I'm doing a multinomial logistic regression.My dependent variable has 3 category.
Percentage distribution for these categories are 5%,72% & 23%.When I'm doing the multinomial,always it is able to predict 95%-98% for second category,4%-5% for 3rd category and 0% for first category. I'm not able to improve prediction accuracy.Is there any other way to do logistic regression with independent variable having skewed in one direction?