A distribution-free convolution model for background correction of oligonucleotide microarray data

BMC Genomics

Table 1 Results of the tests of normality of the background noise as estimated by the exponential-normal convolution model.

Data Set	Test	Rejection Rate	Med P-value	Max P-value
HG-U95Av2	AD	962	0.0008	0.4738
	KS	796	0.0082	0.8261
	SW	999	0.0036	0.1186
HG-U133A	AD	850	0.0064	0.5915
	KS	594	0.0307	0.7700
	SW	962	0.0031	0.3010
GoldenSpike	AD	885	0.0035	0.3559
	KS	639	0.0259	0.7490
	SW	987	0.0016	0.1502

The rejection rates (number of p-values less than 0.05) are much higher than expected, indicating that the background noise is not likely to be normally distributed.
a P-values are not identically 0, but are 0 to at least five decimal places.
AD = Anderson-Darling test, KS = Kolmorgorov-Smirnov test, SW = Shapiro-Wilk test.

ISSN: 1471-2164