From: Applications of Bayesian shrinkage prior models in clinical research with categorical responses
Prediction | |||||||
Priors | MS1 | MS2 | MS3 | MS4 | MS5 | MS6 | MS7 |
N,P,J,m | 400,4, 3, (-2,-1,0,1,2)’ | 250,4,3,(-2,-1,0,1,2)’ | 400,10, 3,0 | 1000,30, 3,0 | 500,50, 3,0 | 600,20, 5,0 | 300,400, 3,0 |
β | (β1,β2,β3)′,β1=(−1.3,1.2,0,0,2)′,β2=(2,−1.5,0,0)′,β3=0 | β1=(−1.3,1.2,0,0,2)′,β2=(2,−1.5,0,0)′,β3=0 | β1 = β2 = (-1.3,-1.2,-0.5,-0.5,2,0,0,0,0,0)’, β3=0 | β1=(A7,B13,C7,D3)′,A7∼N(2,0.25),B13=0,C7∼N(2,0.25),D3=0.β2=(E10,F7,G13),E10=0,F7∼N(2,0.25),G13=0 | β∼N(0,1) | β1=(A7,B13,C7,D3)′,A7∼N(2,0.25),B13=0,C7∼N(2,0.25),D3=0,β2=(E10,F7,G13),E10=0,F7∼N(2,0.25),G13=0 | \(\beta =\left (\begin {array}{l} {\underbrace {N(0,1),\dots,N(0,1)}_{30}},{\underbrace {0,\dots,0}_{170}},\\ {\underbrace {N(0,1),\dots,N(0,1)}_{30}},{\underbrace {0,\dots,0}_{170}} \end {array}\right)\) |
Accuracy | |||||||
Horseshoe | 0.812(0.048) | 0.812(0.047) | 0.712(0.078) | 0.873(0.022) | 0.816(0.044) | 0.760(0.034) | 0.576(0.071) |
Dirichlet Laplace | 0.811(0.048) | 0.811(0.050) | 0.722(0.050) | 0.875(0.023) | 0.818(0.043) | 0.759(0.034) | 0.602(0.068) |
Double Pareto | 0.813(0.048) | 0.812(0.046) | 0.713(0.078) | 0.873(0.021) | 0.816(0.044) | 0.760(0.034) | 0.583(0.069) |
Miss-classification Error | |||||||
Horseshoe | 0.188(0.048) | 0.188(0.047) | 0.282(0.052) | 0.127(0.022) | 0.184(0.044) | 0.240(0.034) | 0.424(0.071) |
Dirichlet Laplace | 0.189(0.048) | 0.189(0.050) | 0.278(0.050) | 0.125(0.023) | 0.182(0.043) | 0.241(0.034) | 0.398(0.068) |
Double Pareto | 0.187(0.048) | 0.188(0.046) | 0.281(0.052) | 0.127(0.021) | 0.184(0.044) | 0.240(0.034) | 0.417(0.069) |
C-Entropy | |||||||
Horseshoe | 0.487(0.071) | 0.491(0.069) | 0.670(0.073) | 0.327(0.031) | 0.432(0.072) | 0.632(0.060) | 1.557(0.308) |
Dirichlet Laplace | 0.480(0.089) | 0.488(0.097) | 0.665(0.100) | 0.313(0.056) | 0.620(0.177) | 0.677(0.101) | 4.113(0.969) |
Double Pareto | 0.488(0.070) | 0.492(0.069) | 0.671(0.072) | 0.330(0.030) | 0.430(0.073) | 0.634(0.059) | 1.963(0.424) |
AUC | |||||||
Horseshoe | 0.719(0.060) | 0.731(0.065) | 0.710(0.047) | 0.877(0.029) | 0.824(0.050) | 0.771(0.030) | 0.632(0.066) |
Dirichlet Laplace | 0.708(0.059) | 0.713(0.072) | 0.716(0.047) | 0.884(0.026) | 0.827(0.048) | 0.767(0.039) | 0.654(0.062) |
Double Pareto | 0.720(0.059) | 0.730(0.064) | 0.713(0.048) | 0.877(0.028) | 0.825(0.048) | 0.773(0.031) | 0.636(0.065) |
Variable Selection | |||||||
Accuracy | |||||||
Horseshoe | 0.996(0.021) | 0.970(0.057) | 0.901(0.040) | 0.986(0.015) | 0.776(0.033) | 0.754(0.039) | 0.850(0.071) |
Dirichlet Laplace | 0.969(0.060) | 0.934(0.091) | 0.929(0.055) | 0.954(0.026) | 0.812(0.039) | 0.789(0.048) | 0.880(0.006) |
Double Pareto | 0.992(0.030) | 0.976(0.052) | 0.903(0.036) | 0.986(0.014) | 0.776(0.034) | 0.749(0.039) | 0.868(0.004) |
Sensitivity | |||||||
Horseshoe | 0.998(0.025) | 0.950(0.101) | 0.838(0.056) | 1.000(0.000) | 0.675(0.045) | 0.659(0.057) | 0.052(0.021) |
Dirichlet Laplace | 0.970(0.082) | 0.920(0.132) | 0.936(0.069) | 1.000(0.000) | 0.849(0.046) | 0.779(0.062) | 0.237(0.033) |
Double Pareto | 0.998(0.025) | 0.965(0.087) | 0.838(0.053) | 1.000(0.000) | 0.677(0.047) | 0.651(0.055) | 0.122(0.026) |
Specificity | |||||||
Horseshoe | 0.995(0.035) | 0.990(0.049) | 0.964(0.056) | 0.978(0.024) | 0.920(0.036) | 0.875(0.051) | 0.000(0.000) |
Dirichlet Laplace | 0.968(0.084) | 0.948(0.125) | 0.922(0.089) | 0.929(0.040) | 0.759(0.063) | 0.803(0.075) | 0.993(0.004) |
Double Pareto | 0.988(0.055) | 0.988(0.055) | 0.968(0.053) | 0.979(0.022) | 0.919(0.040) | 0.875(0.049) | 0.999(0.002) |
L1 error | |||||||
Horseshoe | 0.255(0.035) | 0.294(0.038) | 0.275(0.024) | 0.404(0.012) | 0.335(0.013) | 0.323(0.019) | 0.006(0.006) |
Dirichlet Laplace | 0.215(0.077) | 0.293(0.113) | 0.206(0.060) | 0.246(0.047) | 0.550(0.035) | 0.366(0.055) | 0.024(0.024) |
Double Pareto | 0.260(0.035) | 0.295(0.037) | 0.281(0.024) | 0.414(0.012) | 0.328(0.014) | 0.332(0.018) | 0.011(0.011) |
L2 error | |||||||
Horseshoe | 0.313(0.035) | 0.358(0.037) | 0.354(0.023) | 0.582(0.009) | 0.430(0.013) | 0.409(0.018) | 0.007(0.007) |
Dirichlet Laplace | 0.270(0.105) | 0.366(0.149) | 0.256(0.072) | 0.313(0.064) | 0.687(0.044) | 0.464(0.067) | 0.025(0.025) |
Double Pareto | 0.322(0.035) | 0.363(0.037) | 0.364(0.025) | 0.601(0.010) | 0.419(0.015) | 0.420(0.017) | 0.013(0.013) |