Skip to main content

Table 3 Estimates of power and type I error rates of MBMDR-PC with population structure captured by linear, kernel, and ncMCE principal components and MBMDR-GC, with a nominal 0.05 FWER level

From: Performance of model-based multifactor dimensionality reduction methods for epistasis detection by controlling population structure

  MBMDR-GC MBMDR-PC
  Linear PCA Kernel PCA ncMCE
Sample sizes 200 500 200 500 200 500 200 500
Type I
 Error Rates 0.077 0.085 0.052 0.054 0.054 0.050 0.047 0.055
Power
 Model 1 0.945 0.735a 0.927 0.730a 1.000 0.740a 0.929 0.770a
 Model 2 0.805 0.131a 0.846 0.404a 0.864 0.560a 0.895 0.556a
 Model 3 0.653 0.954 0.658 1.000 0.731 0.970 0.669 0.968
 Model 4 0.483 0.956 0.481 0.960 0.476 0.945 0.490 0.960
 Model 5 0.074 0.784 0.259 0.950 0.238 0.950 0.258 0.958
 Model 6 0.161 0.918 0.447 0.970 0.421 0.970 0.446 0.956
  1. aThe values are close to 1 when the power is calculated based on all significant interactions that include the true signal
\