“…Additionally the effect of different prior distributions for the parameters of the model was analyzed in order to evaluate how prior beliefs could affect the posterior estimates of TP, Se and Sp. Seven scenarios were built assuming prior different beliefs about the parameters (Table 1): (Scenario 1) No prior constraints: TP ∼ Uniform (0, 1), Se ∼ Uniform (0, 1) and Sp ∼ Uniform (0, 1); (Scenario 2) Prior distributions for Se and Sp uniformly distributed according to ranges suggested by Norby et al (2004), and expert opinion of veterinaries in slaughterhouses TP ∼ Uniform (0, 1), Se ∼ Uniform (0.51, 0.98) and Sp ∼ Uniform (0.80, 0.90); (Scenario 3) True prevalence constrained uniformly to values less that 10% and Se and Sp with similar values to previous experiences TP ∼ Uniform (0, 0.10), Se ∼ Uniform (0.50, 1.00) and Sp ∼ Uniform (0.70, 1.00); (Scenario 4) True prevalence constrained uniformly to values less that 5% and Se and Sp with similar values to previous experiences but widen: TP ∼ Uniform (0, 0.05), Se ∼ Uniform (0.40, 1.00) and Sp ∼ Uniform (0.60, 1.00); (Scenario 5) Assuming that Se of the necropsy is 70% and Sp 80% but with high uncertainty TP ∼ Uniform (0, 1.00), Se ∼ Beta (8.00, 4.00) and Sp ∼ Uniform (9.00, 2.00); Scenario 6) Assuming that Se of the necropsy is 70% and Sp 80% but with low uncertainty TP ∼ Uniform (0,1), Se ∼ Beta (71.0, 31.0) and Sp ∼ Uniform (86.0, 16.0); and finally (Scenario 7) TP ∼ Uniform (0, 1), Beta (5.28, 2.07) and Beta (21.20, 2.06) which correspond to our beliefs about necropsies. The agreement among the different laboratory-based diagnostic methods applied in this study was assessed by Cohen kappa statistic under cluster sampling (svykappa function) under R environment, described by Lumley (2004), which is the approach used to asses different tests without assuming that one is the best.…”