Skip to main content

Table 1 Detecting heteroscedasticity

From: Exploratory differential gene expression analysis in microarray experiments with no or limited replication

Dataset

X min

Rho1

Rho1 p-value

Rho2

Rho2 p-value

mac1

-1.50

-0.08

1.36E-03

0.16

0.00E+00

yer033c

-0.70

-0.10

2.30E-08

0.11

3.70E-09

cup5

-0.93

-0.17

0.00E+00

0.11

5.44E-10

spf1

-1.44

-0.18

0.00E+00

0.20

0.00E+00

ymr031c

-0.89

-0.17

0.00E+00

0.11

3.43E-10

vm8

-0.82

-0.10

7.98E-09

0.10

1.66E-08

yap1

-0.97

-0.23

0.00E+00

0.12

1.59E-10

sod1

-1.54

-0.09

3.68E-04

0.15

0.00E+00

fre6

-1.77

-0.13

2.69E-10

0.06

2.46E-04

cin5

-0.28

-0.12

2.17E-14

0.11

1.09E-07

  1. Use of Spearman rank correlation for absolute residuals to detect heteroscedasticity [25] in ten datasets from Hughes et al. [18]. Empirical hyperbolas (here they are based on supsmu smoother) have minima around sample means. As a result, we use two subintervals to compute Spearman rank correlation coefficient: from minus infinity to X min (log2(Cy3) axis) and from X min to plus infinity. We note that sign of Spearman rank correlation always coincides with the sign of first derivative for empirical hyperbolas at a given subinterval (compare Figure 20). Rho1, Spearman coefficient of rank correlation for the former subinterval; Rho1 p-value, p-values for values in column Rho1; Rho2, Spearman coefficient of rank correlation for the latter subinterval; Rho2 p-value, p-values for values in column Rho2 (p-values are given in scientific notation, 0.00E+00 means that the respective p-value was less than 10-16).