Table 7.

True-Positive Rates of Four XGBoost Models on the Test Data Set at Two Different Levels of False-Positive Rates (0.01 and 0.05).

Model	Cutoff score	No. of falsely identified examinees (false-positive rate)	No. of truly identified examinees (true-positive rate)	Precision
False-positive rate of 0.01
Dichotomous responses	0.053	7 (1.1%)	4 (23.5%)	36.3%
Nominal responses	0.052	7 (1.1%)	6 (35.3%)	46.1%
Dichotomous responses + Response time	0.043	7 (1.1%)	7 (41.1%)	50.0%
Nominal responses + Response time	0.078	7 (1.1%)	10 (58.8%)	58.8%
False-positive rate of 0.05
Dichotomous responses	0.010	32 (5.0%)	5 (29.4%)	13.5%
Nominal responses	0.021	32 (5.0%)	8 (47.0%)	20.0%
Dichotomous responses + Response time	0.014	32 (5.0%)	12 (70.5%)	35.3%
Nominal responses + Response time	0.032	32 (5.0%)	13 (76.5%)	28.8%