hr.comp2 {survcomp} | R Documentation |
This function compares two hazard ratios from their betas and standard errors as computed by a Cox model for instance. The statistical test is a Student t test for dependent samples. The two hazard ratios must be computed from the same survival data.
hr.comp2(x1, beta1, se1, x2, beta2, se2, n)
x1 |
risk score used to estimate the first hazard ratio. |
beta1 |
beta estimate for the first hazard ratio. |
se1 |
standard error of beta estimate for the first hazard ratio. |
x2 |
risk score used to estimate the second hazard ratio. |
beta2 |
beta estimate for the second hazard ratio. |
se2 |
standard error of beta estimate for the first hazard ratio. |
n |
number of samples from which the hazard ratios were estimated. |
The two hazard ratios must be computed from the same samples (and corresponding survival data). The function uses a Student t test for dependent samples.
p.value |
p-value from the Student t test for the comparison beta1 > beta2 (equivalently hr1 > hr2) |
hr1 |
value of the first hazard ratio |
hr2 |
value of the second hazard ratio |
Benjamin Haibe-Kains
Student 1908) "The Probable Error of a Mean", Biometrika, 6, 1, pages 1–25.
Haibe-Kains, B. and Desmedt, C. and Sotiriou, C. and Bontempi, G. (2008) "A comparative study of survival models for breast cancer prognostication based on microarray data: does a single gene beat them all?", Bioinformatics, 24, 19, pages 2200–2208.
require(survival) set.seed(12345) age <- as.numeric(rnorm(100, 50, 10) >= 50) size <- as.numeric(rexp(100,1) > 1) stime <- rexp(100) cens <- runif(100,.5,2) sevent <- as.numeric(stime <= cens) stime <- pmin(stime, cens) coxm1 <- coxph(Surv(stime, sevent) ~ age) coxm2 <- coxph(Surv(stime, sevent) ~ size) hr.comp2(x1=age, beta1=coxm1$coefficients, se1=drop(sqrt(coxm1$var)), x2=size, beta2=coxm2$coefficients, se2=drop(sqrt(coxm2$var)), n=100)