Table 2.

Simulation results for sensitivity and specificity with incorrectly specified imperfect reference standards. The random effects of the true models and working models always follow mixture normal distribution and normal distribution, respectively. The diagnostic accuracy of the imperfect reference was generated from (7) using different values of γ₂’s with _D and the diagnostic γ₀ = −4.5, γ₁ = 0.1. The corresponding posterior disease status $S_{1\mid t_i}^D$ accuracy parameter estimates are presented. The true imperfect reference standard was generated from equations (8) and (7) with γ₀ = −4.5, γ₁ = 0.1, and γ₂ = 0.2. The true sensitivity, specificity and disease prevalence were set to be S_e = 0.90, S_p = 0.90, and π₁ = 0.7, respectively.

Number of tests	γ2	Posterior density ω1					Ŝe(se)	Ŝp(se)	π̂1(se)
		$S_{1\mid 0}^D$	$S_{1\mid 1}^D$	$S_{1\mid 2}^D$	$S_{1\mid 3}^D$	$S_{1\mid 4}^D$
5	0	0.02	0.65	0.70	0.74	1.00	0.90(0.065)	0.90(0.067)	0.69(0.050)
	0.1	0.02	0.46	0.70	0.86	1.00	0.90(0.071)	0.90(0.066)	0.71(0.048)
	0.15	0.02	0.36	0.70	0.90	1.00	0.91(0.061)	0.92(0.069)	0.70(0.047)
	0.25	0.02	0.20	0.70	0.95	1.00	0.90(0.060)	0.90(0.058)	0.69(0.057)
	0.3	0.02	0.15	0.70	0.97	1.00	0.91(0.071)	0.91(0.064)	0.70(0.049)
10	0	0.02	0.65	0.70	0.74	1.00	0.89(0.061)	0.90(0.045)	0.70(0.050)
	0.1	0.02	0.46	0.70	0.86	1.00	0.90(0.051)	0.90(0.054)	0.71(0.041)
	0.15	0.02	0.36	0.70	0.90	1.00	0.90(0.059)	0.89(0.051)	0.70(0.037)
	0.25	0.02	0.20	0.70	0.95	1.00	0.89(0.056)	0.91(0.048)	0.70(0.057)
	0.3	0.02	0.15	0.70	0.97	1.00	0.90(0.056)	0.90(0.063)	0.71(0.052)