Table 15.

Incremental validity estimates for the TAPAS composite scores over AFQT for predicting cumulative attrition through 24 months of service.

Attrition Measure/Model	ORAFQT (CI)	ORTAPAS (CI)	rpb	Δ-2LL
6 Month	n = 148,319– 216,372
Can-Do
AFQT	.814 (.800-.828)		.06
AFQT + TAPAS	.803 (.788-.819)	1.002 (1.001–1.003)	.06	12.53
Will-Do
AFQT	.818 (.806-.830)		.06
AFQT + TAPAS	.826 (.815-.839)	.990 (.989-.991)	.08	750.04
Adaptation
AFQT	.818 (.806-.830)		.06
AFQT + TAPAS	.839 (.827-.852)	.991 (.990-.992)	.08	606.08
12 Month	n = 124,391– 191,856
Can-Do
AFQT	.842 (.828-.856)		.05
AFQT + TAPAS	.833 (.818-.849)	1.001 (1.000–1.002)	.05	8.30
Will-Do
AFQT	.850 (.838-.861)		.05
AFQT + TAPAS	.858 (.846-.870)	.990 (.989-.990)	.09	887.69
Adaptation
AFQT	.850 (.838-.861)		.05
AFQT + TAPAS	.873 (.861-.885)	.990 (.990-.991)	.08	743.64
24 Month	n = 88,111– 151,449
Can-Do
AFQT	.822 (.808-.836)		.07
AFQT + TAPAS	.816 (.800-.831)	1.001 (1.000–1.002)	.07	4.05
Will-Do
AFQT	.830 (.820-.841)		.07
AFQT + TAPAS	.838 (.827-.849)	.992 (.991-.992)	.09	614.39
Adaptation
AFQT	.830 (.820-.841)		.07
AFQT + TAPAS	.851 (.840-.862)	.991 (.991-.992)	.10	670.08

OR = odds ratio for each predictor. CI = 95% confidence interval of the odds ratio. r_pb = point biserial correlation between the observed outcome and predicted probability. Δ-2LL = change in negative two log likelihood (deviance) from adding the selected TAPAS composite score to the AFQT-only logistic regression model. Odds ratios equal to 1.0 (or confidence intervals of the odds ratio that include 1.0) indicate no relationship between the predictor and criterion. Odds ratios less than 1.0 indicate a negative relationship between the predictor and criterion. Odds ratios greater than 1.0 indicate a positive relationship between the predictor and criterion. For Δ-2LL, bolded values indicate significant change in model fit based on a Likelihood Ratio χ² test, p < .05.