Skip to main content

Table 2 A summary table of three simulation scenarios

From: Super-delta: a new differential gene expression analysis procedure with robust data normalization

 

Classical

 

Super-delta

 

Oracle

Global

medIQR

Quantile

Cyclic-loess

Mean

Median

MFTM

SIM1

        

Power

88.82(1.02)

88.17(1.08)

87.84(0.99)

87.71(1.00)

87.61(1.01)

88.48(1.04)

88.89(1.13)

88.85(1.00)

Type I error

0.46(0.07)

0.49(0.08)

0.50(0.07)

0.51(0.08)

0.52(0.07)

0.44(0.08)

0.41(0.08)

0.40(0.07)

SIM2

        

Power

92.11(0.81)

90.94(0.89)

90.37(0.95)

90.17(0.93)

89.94(0.95)

91.26(1.01)

91.79(0.94)

92.09(0.77)

Type I error

0.93(0.14)

1.08(0.17)

1.25(0.22)

1.26(0.19)

1.36(0.23)

1.03(0.22)

0.85(0.15)

0.83(0.15)

SIM3

        

Power

89.18(1.49)

76.67(1.51)

77.28(1.84)

76.20(1.63)

76.51(1.59)

77.16(1.89)

86.05(1.70)

88.55(1.62)

Type I error

0.61(0.11)

1.53(0.19)

1.42(0.20)

1.55(0.20)

1.52(0.20)

1.46(0.26)

0.61(0.14)

0.54(0.12)

  1. Sample size is 50 for both groups. All p-values are Benjamini-Hochberg adjusted; Power: Approximate statistical power; Type I error: Approximate Type I error rate. All these measurements are calculated by averaging over 50 replicates. Numbers within parentheses are standard deviations. All numbers are pertentage rates