Table 4

Results of the simulation study: Differences between the p-values of the original two-part test and the two-part permutation test for data sets with small p-values (i.e. p-value of the original test ≤ 0.1), a positive difference means that the two-part permutation test has a smaller p-value than the original test; p1 and p2 are the probabilities for zero values and the positive values in group 1 are shifted by μ; 5,000 data sets with n1 = n2 = 10 were generated for each configuration

Configuration
Number of data sets
Mean difference (± SD)
Median difference
Quartiles of the difference

p1 = p2 = 0, μ = 2.5
4538
0.0118 (± 0.0145)
0.0057
0.0013, 0.0161
p1 = p2 = 0.3, μ = 2.5
4038
0.0008 (± 0.0069)
0.0020
0.0008, 0.0033
p1 = 0.4, p2 = 0.2, μ = 2.5
4346
0.0010 (± 0.0059)
0.0019
0.0008, 0.0033
p1 = 0.2, p2 = 0.4, μ = 2.5
4270
0.0011 (± 0.0056)
0.0018
0.0006, 0.0030
p1 = 0, p2 = 0.3, μ = 2.5
4883
0.0039 (± 0.0055)
0.0020
0.0005, 0.0051
p1 = 0.3, p2 = 0, μ = 2.5
4893
0.0040 (± 0.0053)
0.0022
0.0007, 0.0049
p1 = 0, p2 = 0.4, μ = 0
3427
-0.0033 (± 0.0119)
0.0006
-0.0092, 0.0029
cutoff value = 0.5, μ = 2.5
4621
0.0059 (± 0.0081)
0.0034
0.0009, 0.0072
cutoff value = 1, μ = 2.5
4860
0.0017 (± 0.0052)
0.0013
0.0003, 0.0032

Neuhäuser et al. BMC Bioinformatics 2005 6:35   doi:10.1186/1471-2105-6-35