(a) (b) (c)

Fig. 6.18. Three genes in a simulated toy data.

The experiment to show how the outlier has caused misleading outcome using

where E1, E2 and E3 stand for the prediction errors of the non-DEGs, the up-

DEGs and the down-regulated DEGs. The blue triangles in the curves stand for

rs and the red crosses in the curves stand for the E3 errors.

rediction errors were not significantly reduced using the modified

myth, 2004]. This was not a surprise because the modified t test

ms do not address the outlier problem in data. Using the limma

from zero to ten up-regulated DEGs were mis-classified and from

n down-regulated DEGs were misclassified. Very similar to the t

t, the limma model did not work well when the outlier number

ased. There was no prediction error for the non-DEGs, i.e., hence

I error for using the limma package as well.