Detecting disease-associated genes with confounding variable adjustment and the impact on genomic meta-analysis: With application to major depressive disorder

Table 4 Evaluation of t-test, FEM_minP, FEM_BIC and FEM_ALL methods by simulations

	Type I error (s.e.)				Power (%) (s.e)				Number of DE genes (s.e)				# of variables in Z selected
Scenario	A	B	C	D	A	B	C	D	A	B	C	D	B	C
I	0.051 (.001)	0.046 (.001)	0.049 (.001)	0.051 (.001)	67.9 (.006)	72.9 (.007)	74.6 (.006)	69.7 (.006)	12.5 (1.03)	20.4 (1.11)	23.3 (1.04)	17.6 (1.09)	0.97/1.97*	0.78/1.21*
II	0.051 (.001)	0.052 (.001)	0.050 (.001)	0.051 (.001)	93.8 (.003)	92.9 (.003)	92.5 (.003)	85.0 (.005)	73.4 (.85)	73.0 (.92)	69.8 (.96)	49.7 (1.37)	1.7	0.59
III	0.051 (.001)	0.053 (.001)	0.051 (.001)	0.051 (.001)	93.8 (.003)	92.5 (.004)	91.6 (.004)	85.1 (.005)	71.8 (.93)	68.3 (.94)	66.5 (.88)	45.8 (1.05)	1.8	0.6

*The denominator showed average number of variables in Z selected. The numerator showed average number of selected variables that belong to the true confounders (z₁, z₂).
The average of type I errors, average of statistical powers, and average number of detected DE genes by each method are shown. Standard errors are shown in parentheses. In the last two columns, the average numbers of confounding variables selected by FEM_minP and FEM_BIC are shown.
A: t-test, B: FEM_minP, C: FEM_BIC, D: FEM_ALL

ISSN: 1471-2105