was composed of 162 likely missing DEGs or the Type II errors

e low extreme outliers occurred in the cancer replicates.

A low extreme outlier in gene 218765_at in GDS3139 caused a Type II error.

ure 6.15, a high extreme outlier presented in the cancer replicates.

of this outlier, the mean expression of the cancer replicates was

p. This certainly helped enlarge the mean difference between two

ns. Therefore the raw p value was 0.0098 if the outlier was not

However if the outlier was removed, the new p value became

his shows that the occurrence of a high extreme outlier in the

plicates caused a false DEG, a Type I error. If this high extreme

as removed, the gene becomes a non-DEG. The investigation in

e data set found 26 likely false DEGs (the Type I errors) in total

e high extreme outliers presented in the cancer replicates.

A high extreme outlier in gene 212954_at in GDS3139 caused a false DEG.

sing DEG or a Type II error may happen due to a high extreme

curred in the cancer replicates as shown in Figure 6.16. This high

outlier was far away from the cluster of the rest cancer replicates

above the control expression cluster. The raw p value was 0.1188