Compare groups
Test treatment differences using death and possibly recurrent nonfatal events, with optional stratification.
WRrec()
Two-sample vignette Clinical priorities
guide the comparison.
Analyze death and nonfatal events together, with their clinical importance built into the method.
The principle
Compare survival first
A hospitalization does not automatically outweigh a longer life.
Hospitalization End of follow-up
A wins on survivalNo further comparison is needed.
In practice
# Load the included trial subset
library(WR)
dat <- hfaction_cpx9
# Stratify by age
fit <- WRrec(
ID = dat$patid,
time = dat$time,
status = dat$status,
trt = dat$trt_ab,
strata = dat$age60
)
print(fit)# Load the regression dataset
library(WR)
dat <- non_ischemic
Z <- as.matrix(dat[, c("trt_ab", "age")])
# Adjust for age
fit <- pwreg(
ID = dat$ID,
time = dat$time,
status = dat$status,
Z = Z
)
print(fit)The default is the last-event-assisted win ratio.
This example compares exercise training with usual care in the included high-risk HF-ACTION subset, stratified by age. See the full analysis ↗
Choose your analysis
Adjust for covariates
Fit proportional win-fractions regression models, including stratified models, and examine standardized score processes.
pwreg() · score.proc()
Regression vignette Plan a study
Calculate sample size for the standard win ratio test using outcome assumptions, accrual, follow-up, and target power.
WRSS()
Sample size vignette A useful distinction
WRrec() can use recurrent nonfatal events. Standard pwreg() uses the first nonfatal event. Choose the analysis to match the scientific question.
Get started
Install WR
Install WR from CRAN, then try the included HF-ACTION data.
install.packages("WR")
library(WR)Data structure
Prepare data in long format
Use one row per event or censoring record, with a patient identifier and event time.
- 0
- Censoring
- 1
- Death
- 2
- Nonfatal event
Go deeper
Documentation and papers
Read the assumptions, understand the data structure, and find the right functions for your study.