Other steps
have priority. The correct answer is either (b) or (c). A thorough
study of the clinical problem is required before a valid predictive
model can be developed, and by such study, a list of candidate
predictors can be assembled and the relevance of the data set
judged.