. Author manuscript; available in PMC: 2014 Dec 26.

Published in final edited form as: Stat Methods Med Res. 2013 Jun 26;25(4):1596–1619. doi: 10.1177/0962280213492588

Table 4.

GLMM method estimates and standard errors (SE)

Model	Sensitivity	Specificity	ρ	AUC	σ_μ(SE)	σ_ν(SE)	AIC	-2logL
logit link	0.827 (0.042)	0.888 (0.021)	−0.298 (0.330)	0.908 (0.049)	1.143 (0.243)	0.679 (0.208)	214.8	204.8
logit link-independence	0.826 (0.042)	0.887 (0.021)	0	0.902 (0.026)	1.138 (0.241)	0.678 (0.206)	213.5	205.5
probit link	0.817 (0.042)	0.885 (0.021)	−0.312 (0.325)	0.915 (0.051)	0.636 (0.132)	0.359 (0.111)	215.2	205.2
c-log-log link	0.801 (0.045)	0.882 (0.021)	−0.329 (0.321)	0.925 (0.041)	0.560 (0.119)	0.284 (0.090)	215.9	205.9

AUC denoted the area under the summary ROC curve. The boldfaced cells represent the best chosen model based on AIC. The ‘logit link-independece’ model uses logit link and assumes independence between sensitivity and specificity while the other models assume dependence.