TABLE 5.

Measures of Accuracy for Four Final Statistical Models Predicting MLD, Using Item Scores as Predictors, and MLD in Grades 2 and 3 as the Outcome Variable, n = 207

	Model 4	Model 5	Model 6	Model 7
Items in model	8	6	4	17
Sensitivity	79.0%	81.8%	82.6%	71.4%
Specificity	88.0%	90.3%	83.9%	84.5%
PPV	40.5%	50.0%	38.8%	26.3%
NPV	97.6%	97.7%	97.5%	97.5%
Correctly classified	87.1%	89.4%	83.7%	83.6%
ROC	0.89	0.87	0.88	0.88

Note. PPV = postive predictive value; NPV = negative predictive value; ROC = receiver operating curve.

Item scores included the four commonly occurring TEMA-2 items that appeared in the most accurate of our first set of analyses, and which appear in Table 7. Models 4 and 5 included two additional items from either the TEMA-2 or the KeyMath-R, respectively; whereas Model 6 included only the four core TEMA-2 items. Model 7 included 13 more items from the TEMA-2 and KeyMath-R.