Plot the hazard ratio of a covariate as a function of landmark time

Plot the hazard ratio of a covariate as a function of landmark time

Usage

plot_hazard_ratio(
  x,
  covariate,
  landmarks,
  w,
  type = c("crude", "supermodel", "both"),
  contrast = 1,
  newdata = NULL,
  level = 0.95
)

Arguments

x

An object of class LandmarkAnalysis.

covariate

Character name of the (numeric) covariate to compute the hazard ratio for.

landmarks

Numeric vector of landmark times. For type = "crude", a fit must already exist in x@survival_fits for each landmark (via fit_survival) at window w; for type = "supermodel", a superlandmark model must already exist for this landmarks/w pair (via fit_superlandmark).

w

Scalar prediction window, as passed to fit_survival (horizons landmarks + w) and/or fit_superlandmark.

type

Whether to plot the crude per-landmark estimates, the pooled supermodel curve, or both.

contrast

Numeric increase in covariate to compute the hazard ratio for (defaults to 1, i.e. the usual per-unit hazard ratio).

newdata

A single-row data.frame giving values for every other covariate in the model (excluding covariate and LM). Defaults to the mean/first level of each from x@data_static.

level

Confidence level for the pointwise intervals (default 0.95).

Value

A ggplot object.

Details

For every landmark time, the hazard ratio associated with a contrast-unit increase in covariate is computed as \(\log(HR) = (x_1 - x_0)^\top \hat\beta\).

Any other covariate present in the formula is held fixed at the mean (numeric) or first level (factor) of x@data_static, unless overridden via newdata.