If the data doesn't match your bias; change your definitions and re-run your study.