Skip to main content
. 2009 Dec 23;26(4):501–508. doi: 10.1093/bioinformatics/btp707

Table 4.

Case III: performance of PMFA, MFA and PMND for five simulation set-ups

Method Set-up 0
Set-up 1
Set-up 2
Set-up 3
Set-up 4
μ2 = 0, c = 0
μ2 = 4.5, c = 1
μ2 = 4.5, c = 2
μ2 = 6.0, c = 1
μ2 = 6.0, c = 2
ĝ N z1 z2 N z1 z2 N z1 z2 N z1 z2 N z1 z2
1 50 20 80 4 0 34.8
2 20 0 46.8 37 0 52.9 29 0 51.8 26 0 53.1
PMFA 3 26 0 57.9 13 0 58.5 21 0 59.5 24 0 67.7
4
5
1 50 0 0 49 0 0 49 0 0 45 0 0 41 0 0
2 1 0 0 1 0 0 5 0 0 9 0 0
MFA 3
4
5
1 50 20 80
2
PMND 3
4
5 50 0 35.3 50 0 30.4 50 0 28.8 50 0 53.4

Among K = 100 variables, K1 = 20 were informative; among n = 100 observations, n1 = 60 were in one cluster. N represents the number of datasets identified with ĝ clusters; z1 and z2 represent the average number of deleted informative and noise variables, respectively, among datasets identified with ĝ clusters.