Correlation versus curve fitting in research on accident proneness: reply to Martiz.