A comparative study of forest methods for time-to-event data: variable selection and predictive performance

Table 1 Information of each simulated dataset

model	form	covariate
A1	1.5x_1i + Ι(x_5i = 2)	x_1i − x_2i~U(0, 1), x_3i − x_4i~Ν(0, 1), x_5i − x_6i~DiscreteU(1, 2), x_7i − x_8i~DiscreteU(1, 4), x_9i − x_10i~DiscreteU(1, 8)^a
A2	1.5x_1i + Ι(x_7i ≥ 3)
A3	1.5x_1i + Ι(x_9i ≥ 5)
B1	Ι(x_1i > 0.5) ∗ Ι(x_5i = 2)
B2	Ι(x_1i > 0.5) ∗ Ι(x_7i ≥ 3)
B3	Ι(x_1i > 0.5) ∗ Ι(x_9i ≥ 5)
C	x_1i + 1.5x_2i	x_1i − x_10i~MVΝ(0, Σ)^b
D1	2x_1i + 3x_2i	x_1i − x_Mi~Ν(0, 1)^c
D2	2x_1i + 3x_2i	x_1i − x_Mi~DiscreteU(1, 2)^a

^a DiscreteU(1, k) is the discrete uniform distribution, a simple distribution that puts equal weight on the integers from 1 to k
^b Σ is a squared matrix with all diagonal elements equal to 1 and all off-diagonal elements equal to ρ
^c M is the number of covariates

ISSN: 1471-2288