. Author manuscript; available in PMC: 2012 Jul 13.

Published in final edited form as: J Consult Clin Psychol. 2004 Oct;72(5):809–820. doi: 10.1037/0022-006X.72.5.809

Table 2. Different Screening Models Predicting to Parent-Teacher Fifth Grade Externalizing Outcome: Base Rate = 20%.

						PPV >.50		Sensitivity >.50

Model no.	Predictor	Likelihood ratio χ²	R²	AIC	Δ	Sensitivity	Specificity	Sensitivity	Specificity
1	Teacher K^*	47.1	.15	240	47	.45	.88	.52	.86
2	Teacher 1st^*	66.4	.21	221	28	.50	.88	.50	.88
3	Teacher K^* & Teacher 1st^*	74.7	.23	214	21	.57	.88	.50	.91
4	Teacher K^* & Parent K^*	59.2	.19	230	37	.48	.88	.50	.87
5	Teacher 1st^* & Parent 1st^*	93.6	.28	195	2	.64	.88	.52	.95
6	Teacher K & Teacher 1st^* and Parent K & Parent 1st	99.9	.30	193	0	.68	.88	.50	.94

Note. Predictors in bold had significant individual χ² values. Test values in bold surpass preset criteria. Models with a delta greater than 4 are notably inferior to the best model (Δ = 0); models with a delta greater than 10 fail to explain a substantial amount of variation in the model. PPV = positive predictive value; AIC = Akaike's information criterion; K = kindergarten; 1st = first grade.

p <.001.