Table 2

Type I error for two typical PEM in different scenarios

Sample size501002003005001000
Pathway length K=2Embedded Image Embedded Image
Multivariate Δ*0.0270.0390.0450.0320.0400.051
Exact*0.0190.0350.0420.0310.0400.050
Unbiased*0.0330.0430.0450.0330.0400.051
Bootstrap*0.0400.0520.0570.0420.0490.059
Percentile bootstrap†0.0460.0570.0590.0470.0570.059
Bias-corrected bootstrap†0.0560.0680.0580.0480.0570.058
Pathway length K=3Embedded Image Embedded Image
Multivariate Δ*0.0050.0210.0370.0350.0480.045
Exact*0.0030.0150.0330.0350.0470.044
Unbiased*0.0150.0260.0420.0400.0500.047
Bootstrap*0.0110.0200.0440.0370.0530.048
Percentile bootstrap†0.0290.0340.0490.0470.0540.050
Bias-corrected bootstrap†0.0590.0570.0590.0530.0580.057
Pathway length K=4Embedded Image Embedded Image
Multivariate Δ*0.0040.0120.0220.0300.0350.047
Exact*0.0010.0090.0210.0260.0320.047
Unbiased*0.0100.0190.0320.0350.0370.047
Bootstrap*0.0140.0200.0340.0370.0390.054
Percentile bootstrap†0.0330.0340.0420.0450.0480.058
Bias-corrected bootstrap†0.0650.0600.0550.0520.0540.056
Pathway length K=5Embedded Image Embedded Image
Multivariate Δ*0.0000.0040.0180.0340.0350.032
Exact*0.0000.0030.0140.0310.0330.029
Unbiased*0.0010.0130.0230.0390.0380.033
Bootstrap*0.0040.0070.0260.0370.0370.041
Percentile bootstrap†0.0220.0350.0430.0490.0490.043
Bias-corrected bootstrap†0.0660.0610.0590.0570.0560.046
  • *For PEM-UD with different methods, estimate the variance.

  • †For PEM-D with different methods, estimate the CI.