Be cautious in interpreting the results of a stepwise procedure, especially in determining which variables were not entered. Were they really unimportant, or were they simply correlated with other variables that did enter into the equation?
Question 4. The researcher reported that 75% of the subjects were correctly classified on the basis of the equation. Can you expect results this good if you now used the equation with your patients?
Answer. No. A discriminant function always does better in "classifying" the subjects from which it was derived than in doing so with a new group. How much better is extremely difficult to determine.
Was this article helpful?