Table 3.

Generalized linear regression model predicting item difficulty from correct probe trials, incorrect probe trials, and the joint model

Predictor	β [95% CI]	SE	z	p
Correct probe trials
(Intercept)	0.839 [0.535, 1.143]	0.155	5.411	< 0.001***
ProbabilityProbe_LastHeard	1.052 [0.364, 1.740]	0.351	2.999	0.003 **
Level	− 0.116 [− 0.169, − 0.063]	0.027	− 4.332	< 0.001***
Incorrect probe trials
(Intercept)	− 1.633 [− 2.303, − 0.963]	0.342	− 4.770	< 0.001***
ProbeTrueIm_AbsDiff	1.028 [0.763, 1.293]	0.135	7.607	< 0.001***
Heard_Range	− 0.224 [− 0.377, − 0.071]	0.078	− 2.886	0.004 **
Joint model
(Intercept)	− 0.918 [− 1.516, − 0.320]	0.305	− 3.009	0.003 **
Incorrect: ProbabilityProbe_LastHeard	0.228 [− 1.277, 1.733]	0.768	0.296	0.767
Correct: ProbabilityProbe_LastHeard	2.778 [1.602, 3.954]	0.600	4.627	< 0.001***
Incorrect: level	− 0.157 [− 0.282, − 0.032]	0.064	− 2.438	0.015 *
Correct: level	− 0.176 [− 0.266, − 0.086]	0.046	− 3.795	< 0.001***
Incorrect: ProbeTrueIm_AbsDiff	1.151 [0.820, 1.482]	0.169	6.823	< 0.001***
Incorrect: Heard_Range	− 0.553 [− 0.763, − 0.343]	0.107	− 5.164	< 0.001***
Correct: Heard_Range	0.157 [0.014, 0.300]	0.073	2.162	0.031 *

In the joint model, correct indicates Probe_Accuracy = 1; incorrect indicates Probe_Accuracy = 0

p values are estimated using Wald tests