In junkka/ehahelper: Helper Functions for Event History Analysis

Event history analysis helper package

Helper functions for event history analysis with the survival package.

Install

Install from GitHub repository.

library(devtools)
install_github('junkka/ehahelper')

Functions

ggsurv

Make a data.frame of a survfit or coxph object for visualization with ggplot2.

library(ehahelper)
library(survival)
library(ggplot2)
surv_object <- coxph(Surv(time, status) ~ strata(x), data = aml)
ggplot(ggsurv(surv_object), aes(time, surv, color=strata)) + geom_step()

gg_zph

ggplot2 implementation for visualizing scaled Schoenfield residuals form a cox.zph object.

bladder1 <- bladder[bladder$enum < 5, ] 
fit <- coxph(Surv(stop, event) ~ (rx + size + number) * strata(enum) + 
          cluster(id), bladder1)
x <- cox.zph(fit, transform = "identity")
gg_zph(x)
gg_zph(x, log = TRUE)

coxme tidyer

Convert coxme objects to tidy format.

library(broom)
library(coxme)
fit <- coxme(Surv(y, uncens) ~ trt + (1|center), eortc)
knitr::kable(tidy(fit, exp = T), digits = 3)

fit_g <- glance(fit)
knitr::kable(as.data.frame(t(fit_g)), digits = 3)

coxme augment

Using a coxme model, add fitted values and standard errors to original dataset.

eortc_augmented <- augment(fit, eortc)

knitr::kable(head(eortc_augmented))

coxme predict

Get predicted vales based on a mixed-effects Cox model, fitted using the coxme package. Extends the standard predict.coxme function by allowing for new data, and by calculating relative risks, either overall or within stratum.

new_data <- data.frame(trt = unique(eortc$trt))

predict_coxme(fit, newdata = new_data, type = "risk")

junkka/ehahelper documentation built on March 17, 2021, 2:12 a.m.