Skip to main content

Table 1 Bias, RMSE, and Coverage Probability of 95% Credible Interval of the BSM, NNT, CC, and MI method, When Sensitivity Parameter is Set to True Value

From: Accounting for misclassification bias of binary outcomes due to underscreening: a sensitivity analysis

    Low-Low Low-High High-Low High-High
BSM Bias*1000 b 1 3.19 4.14 1.91 2.88
b 2 12.94 4.81 2.60 1.88
RMSE*1000 b 1 192.51 163.04 94.14 66.59
b 2 75.19 54.64 40.47 28.12
Coverage Probability b 1 96.0% 97.0% 97.0% 97.5%
b 2 88.0% 94.5% 88.5% 91.0%
NNT Bias*1000 b 1 −309.37 −101.24 −857.09 −331.60
b 2 136.93 56.88 30.70 34.85
RMSE*1000 b 1 362.4 190.85 860.28 338.08
b 2 151.02 77.23 41.93 42.98
Coverage Probability b 1 63.0% 95.0% 0% 0%
b 2 43.5% 86.0% 81.0% 73.5%
CC Bias*1000 b 1 89.30 43.30 113.73 77.02
b 2 −38.4 −24.67 −51.17 −37.12
RMSE*1000 b 1 212.04 167.71 148.09 101.61
b 2 81.55 60.70 64.04 45.83
Coverage Probability b 1 93.0% 96.0% 76.5% 83.5%
b 2 89.5% 91.5% 68.5% 74.5%
MI Bias*1000 b 1 97.01 43.23 115.96 79.41
b 2 −43.32 −25.72 −51.34 −35.10
RMSE*1000 b 1 221.17 168.02 150.53 103.85
b 2 84.14 61.07 64.81 44.29
Coverage Probability b 1 92.5% 97.0% 76.0% 82.0%
b 2 88.0% 92.5% 71.5% 76.0%
  1. *denotes "multiplied by"
  2. BSM Bayesian selection model, NNT Negative for not tested, CC Complete-case analysis, MI Multiple imputation