plot_predicted_probs: Function to plot the predicted probability of an event using...

In benjamin-w-campbell/coxtools: Tools for Cox Proportional Hazards Models

Description

This function assists the user in interpreting effect sizes estimated using the coxph function. It provides the predicted probability of not observing an event, the survival function, for a fixed time period over the full range of variable's values. This routine is a plot-based wrapper for the "riskRegression" "predictCox" function, and as such, should be treated accordingly.

Usage

plot_predicted_probs(coxph_fit = NULL, var = NULL, time = NULL,
  seed = NULL,
  yaxis_label = "Survival Function: Probability of Not Observing Event",
  xaxis_label = "Variable Values", title = NULL)

Arguments

coxph_fit	The output from a fitted "coxph" call.
var	A character string specifying the variable from "coxph_fit" that predicted probabilities should be plotted for.
time	This is the fixed time point that should be used to calculate the predicted probabilities.
seed	This is the seed that should be set for replication
yaxis_label	This is the label as it appears for the y-axis. This is the probability of not observing an event.
xaxis_label	This is the label ax it appears for the x-axis. This is the range of variable values for the "var" variable.
title	This is the plot title as it should appear.

Value

A ggplot2 object summarizing the corresponding predicted probability.

Examples

library(survival)
data("heart")
# Coerce from factor
heart$transplant <- as.numeric(heart$transplant)
# Rescale age
heart$age <- heart$age+48
coxph_fit <- coxph(Surv(start, stop, event) ~
               age + transplant +surgery,
             data = heart,
             x = TRUE)

plot_predicted_probs(coxph_fit,
                     var = "age",
                     time = mean(heart$stop-heart$start),
                     seed = 123,
                     xaxis_label = "Age")

benjamin-w-campbell/coxtools documentation built on May 31, 2019, 10:46 p.m.