Comparison of small n statistical tests of differential expression applied to microarrays

BMC Bioinformatics

Table 4 False positive rates

		t-stat	CyberT	LPE	BRB	limma	FC
HGU133A 22258 genes	MAS 5.0	.064	.078	.060	.063	.062	.05
	RMA	.069	.084	.044	.078	.078	.05
	gcRMA	.050	.079	.031	.060	.061	.05
	dChip	.093	.115	.081	.105	.105	.05
	LMGene	.082	.099	.065	.095	.093	.05
	VSN	.074	.091	.049	.091	.088	.05
HGU95 12614 genes	MAS 5.0	.044	.05	.042	.046	.046	.05
	RMA	.040	.037	.011	.035	.034	.05
	gcRMA	.009	.018	.003	.017	.018	.05
	dChip	.043	.041	.003	.040	.040	.05
	LMGene	.040	.035	.008	.035	.033	.05
	VSN	.040	.035	.008	.035	.033	.05
cDNA liver vs liver 27648 genes	Loess	.002	.001	0	.001	.001	.05
	Loess(BC)	.01	.002	.001	.007	.007	.05
	glog Loess	.007	.002	0	.002	.002	.05
	glog Loess(BC)	.014	.008	.001	.011	.012	.05
	VSN	.006	.001	0	.001	.001	.05
cDNA liver vs pool 27648 genes	Loess	.138	.135	.034	.151	.151	.05
	Loess(BC)	.174	.171	.106	.169	.173	.05
	glog Loess	.233	.238	.151	.251	.252	.05
	glog Loess(BC)	.233	.235	.158	.238	.240	.05
	VSN	.244	.249	.183	.251	.252	.05
Simulated 9900 genes	Common	.048	.048	.032	.046	.047	.05
	Local	.044	.044	.028	.046	.045	.05
	Inverse Gamma	.048	.053	.039	.048	.048	.05

Average false positive rates (using a p value threshold of .05) of the six tests of differential expression. The LMGene package was used to apply the Loess normalization and glog transformation to the cDNA data. BC indicates that background correction was applied to the data before normalization.

ISSN: 1471-2105