plot_calibration: Plot the calibration for MMM

In JanvandenBrand/jmmm: High Dimensional Joint Models

Plot the calibration for MMM

Description

Create a grobs with calibration plot for a given landmark and horizon.

Usage

plot_calibration(
  predictions,
  prediction_landmark,
  prediction_horizon,
  id,
  failure_time,
  failure
)

Arguments

predictions	the predictions returned by the mmm_predictions function
prediction_landmark	a numeric vector with the landmark times to select predictions
prediction_horizon	the prediction horizon at which to calculate the ROC-AUC.
id	a character value for the name of the vector of subject identifiers in the predictions data
failure_time	a character value for the name of the vector with failure times in the predictions data
failure	a character value with the name of the vector of failure indicators.

Value

a list of grobs

Examples

## Not run: 
# Retrieve calibration data for plotting
plot_calibration(
  predictions=predictions,
  prediction_landmark=1,prediction_horizon=horizon,
  id="id",
  failure_time="stime",
  failure="failure"
)

# Create a set of calibration plots from a list of predictions
  plot_calibration <- lapply(which(lm %in% c(1:3)), function(l) {
    plot_calibration(
    predictions=predictions[[l]],
    prediction_landmark=lm[l],
    prediction_horizon=horizon,
    id="id",
    failure_time="ftime",
    failure="failure"
  )
})

# Combine the calibration data
plot_calibration <- do.call(rbind, plot_calibration_train)

# Create the plot with overlayed calibration lines for each landmark
plot_calibration_train <- ggplot(
    plot_calibration_train,
    aes(x=Pred, y=Obs)
  ) +
  geom_smooth(
    aes(color=factor(landmark),
    fill=factor(landmark)),
    method="loess",
    span=0.3,
    alpha=0.2
 ) +
 geom_abline(intercept=0, slope=1, alpha=0.5) +
 geom_rug(aes(x=Pred)) +
 scale_y_continuous(breaks=seq(0, 1, 0.2)) +
 scale_x_continuous(breaks=seq(0, 1, 0.2)) +
 coord_cartesian(
   xlim=c(0, 1),
   ylim=c(0, 1)
 ) +
 abs(title="",
   caption=paste("Calibration plots for kidney graft failure at",
     horizon,
     "years follow-up in training data"),
   y="Observed risks",
   x="Predicted risks",
   fill="Landmark"
 ) +
 guides(color="none") +
 ggthemes::theme_tufte(ticks=TRUE)
 
## End(Not run)

JanvandenBrand/jmmm documentation built on May 30, 2022, 9:37 a.m.