Gaze behavior and cognitive states during fingerprint target group localization

Table 3 Performance of logistic regression models predicting the subphase or location of fixations (n = 22,747 Comparison-phase fixations on the right image from 675 trials). Model performance is compared using area under the (receiver operating characteristic) curve (AUC). No cross-terms were used in these models: each model had one degree of freedom for each predictor term

Model	Cropped 13,770 fixations		Latent and Plain 8794 fixations
Model	Subphase A	Subphase C	Subphase A	Subphase C
Speed3	0.73	0.81	0.80	0.69
PercentOfFixesInCell	0.67	0.84	0.74	0.73
TimeInImage	0.46	0.75	0.60	0.52
DetailedBackAndForth	0.63	0.68	0.65	0.57
Speed3 + PercentOfFixesInCell	0.74	0.87	0.83	0.76
Speed3 + TimeInImage	0.72	0.86	0.79	0.70
Speed3 + DetailedBackAndForth	0.74	0.83	0.81	0.69
TimeInImage + DetailedBackAndForth	0.66	0.79	0.72	0.56
Speed3 + TimeInImage + DetailedBackAndForth	0.74	0.86	0.81	0.70
Speed3 + PercentOfFixesInCell + TimeInImage + DetailedBackAndForth	0.78	0.89	0.84	0.76