Assessing Scientific Practices Using Machine-Learning Methods: How Closely Do They Match Clinical Interview Performance?