(a) (b)

n illustration of Type I and Type II errors when drawing a small number of

from the letrozole drug data. (a) For the gene FKBP15. (b) For the gene

oarray gene expression analysis

nderstood from the above analysis that the t test may not be robust

replicate number is small, a question is how to deal with this

. A direct consideration is to compensate the denominator of the

, or modify it so as to make it suitable to data with insufficient

number. An approach called the significant analysis for

ays (SAM) has been proposed to adjust the denominate of a t

or this purpose. The method is called the modified t test [Tusher,

01]. It has a wide application in gene expression pattern analysis

i, et al., 2018; Tsuda, et al., 2019]. How to improve the

nce of the modified t test still draws the attention in the

ty [Zhang, 2007; Li, et al., 2013; Tzeng, 2021]. The modified t

used in the modified t test is defined as below,

ݐൌ

ݑො஺െݑො஻

ߪሺݔ஺, ݔ஻ሻ൅ߙ

(6.6)

is positive and is the focal point of the research of gene

n analysis for DEG discovery. A p value calculated based on this

t statistic is called a modified p value. When ߙ is introduced, the

c will be reduced. The consequence is a greater p value.

cally, the chance of introducing the Type I error in a DEG

y process is reduced using this modified t test. However, if ߙ is

, the chance of introducing the Type II error will be increased.

accurately estimate ߙ is then a key question of the research. To