Pseudo-Melanoma Survival Data — melanoma • survivalMPL

Simulated partly interval-censored survival data based on the design in Moore (2016, Examples 2.6–2.7). The true baseline hazard is Weibull with shape 0.5: \(h_0(t) = t^{-1/2}\), \(H_0(t) = 2\sqrt{t}\).

Format

A data frame with 300 observations on 10 variables:

t_L: Numeric lower bound of the observed interval (0 for left-censored observations).
t_R: Numeric upper bound of the observed interval (Inf for right-censored observations).
Arm: Indicator: tumour located on the arm (reference: head/neck).
Leg: Indicator: tumour located on the leg.
Trunk: Indicator: tumour located on the trunk.
mm1to2: Indicator: tumour thickness 1–2 mm (reference: < 1 mm).
mm2to4: Indicator: tumour thickness 2–4 mm.
mm4plus: Indicator: tumour thickness > 4 mm.
Female: Indicator: patient is female.
Age_centred: Age in decades, centred at the sample mean.

Source

Moore, D.K. (2016). Applied Survival Analysis Using R. Springer.

Details

Observations are classified as exact events (t_L == t_R), left- censored (t_L == 0), right-censored (t_R == Inf), or interval-censored otherwise. Two thirds of the sample (200 subjects) are exact events and one third is censored: 58 interval censored, 23 right censored and 19 left censored. Pass to coxph_mpl via Surv(t_L, t_R, type = "interval2").

True regression coefficients (beta_true) used in the simulation:

`Arm`	-0.56
`Leg`	0.01
`Trunk`	-0.22
`mm1to2`	0.22
`mm2to4`	0.87
`mm4plus`	1.13
`Female`	-0.17
`Age_centred`	0.14

Generated with set.seed(1) and simulation parameters pi_E = 2/3 (probability of an exact event), a1 = 0.02 and a2 = 0.30 (inspection-window scales); see dev/data-raw/melanoma.R to reproduce.

Examples

data(melanoma)
if (FALSE) { # \dontrun{
fit <- coxph_mpl(
  Surv(t_L, t_R, type = "interval2") ~
    Arm + Leg + Trunk + mm1to2 + mm2to4 + mm4plus + Female + Age_centred,
  data = melanoma, basis = "m"
)
summary(fit)
} # }