Combining techniques for screening and evaluating interaction terms on high-dimensional time-to-event data

BMC Bioinformatics

Table 1 The effect sizes of non-zero effects in each scenario

Scenarios:	Effect size ME	Effect size Int	Corr value	Block
				size
Sim42	(3, 3, -3, -3)	(5,-5)
Sim22_1.0	(0.9, -0.9)	(1.0, -1.0)
Sim22_0.5	(0.9, -0.9)	(0.5, -0.5)
Sim22_0.25	(0.9, -0.9)	(0.25, -0.25)
Sim22_1.5	(0.9, -0.9)	(1.5, -1.5)
Sim22_2.0	(0.9, -0.9)	(2.0, -2.0)
Sim22_2.5	(0.9, -0.9)	(2.5, -2.5)
Sim22_bin	(0.9, -0.9)	(1.0, -1.0)
Sim22_corr01	(0.9, -0.9)	(1.0, -1.0)	0.1	5
Sim22_corr03	(0.9, -0.9)	(1.0, -1.0)	0.3	5
Sim22_corr05	(0.9, -0.9)	(1.0, -1.0)	0.5	5
Sim22_corr07	(0.9, -0.9)	(1.0, -1.0)	0.7	5

ME: main effect. Int: interaction. Corr: correlation. In all scenarios, the samples size is 150 (=n) and the number of covariates is 1000 (=p). The effect sizes are given in the form ‘(coefficient value of effect 1, coefficient value of effect 2)’. For the scenarios with correlations, p divided by the block size (200) gives the dimension of the normal distribution from which the values of the variables in a block are sampled, and the correlation value is the value at the off-diagonals of the corresponding covariance matrix. In scenarios Sim42 and Sim22_1.0, all interaction detection strategies are covered. All other scenarios are used for investigating rsf-VIF-res.

ISSN: 1471-2105