. Author manuscript; available in PMC: 2011 May 20.

Published in final edited form as: Stat Med. 2010 May 20;29(11):1206–1218. doi: 10.1002/sim.3862

Table 4.

Summary of posterior estimates using the Bayesian approach under the assumption of homogenous dependence. The triple notation of _LP_U denotes the posterior median P with 95% equal tailed credible limits (L, U). The posterior estimates of the probabilities of disease classes have been multiplied by 1000 for presentation.

Study	Disease status	Independent Model ( ${\hat{P}}_{d}^{⊥}$ )	Dependent Models
Study	Disease status	Independent Model ( ${\hat{P}}_{d}^{⊥}$ )	the α model ( ${\hat{P}}_{d}^{α}$ )	the θ model ( ${\hat{P}}_{d}^{θ}$ )	the ρ model ( ${\hat{P}}_{d}^{ρ}$ )	BMA*
1	No cancer	₉₉₁995₉₉₇	₈₇₈953₉₈₂	₈₆₂926₉₆₂	₈₈₅992₉₉₆	₈₆₈955₉₉₅
1	cancer	₃5₉	₁₈47₁₂₂	₃₈74₁₃₈	₄8₁₁₅	₅45₁₃₂

2	Not present	₃₄52₉₆	₁₅₃381₆₆₀	₂₁₅418₆₀₈	₅₄255₈₈₃	₆₈383₈₁₆
	Low grade	₈₉₇941₉₅₉	₃₂₁601₈₃₃	₃₂₉520₇₄₆	₁₀₈737₉₃₈	₁₇₅582₉₂₄
	High grade	₅7₁₀	₇15₄₃	₃₀57₉₅	₅8₁₃	₆15₈₄

Note: Study 1, Health Insurance Plan Screening Study for Breast Cancer in New York. Study 2, a multicenter study for screening cervical cancer. BMA = Bayesian model averaging based on the α, θ and ρ dependent models.