Subtyping women with bulimia nervosa along dietary and negative affect dimensions: further evidence of reliability and validity.