ROC: Receiver Operator Characteristics (ROC) for a 'joint' model.
In gmvjoint: Joint Models of Survival and Multivariate Longitudinal Data

View source: R/ROC.R

ROC	R Documentation

Receiver Operator Characteristics (ROC) for a `joint` model.

Description

Using longitudinal information available up to a time, establish diagnostic capabilities (ROC, AUC and Brier score) of a fitted joint model.

Usage

ROC(fit, data, Tstart, delta, control = list(), progress = TRUE, boot = FALSE)

Arguments

`fit`	a joint model fit by the `joint` function.
`data`	the data to which the original `joint` model was fit.
`Tstart`	The start of the time window of interest, `Tstart` denotes the time point up to which longitudinal process(es) is used in calculation of survival probabilities.
`delta`	scalar denoting the length of time interval to check for failure times.
`control`	list of control arguments to be passed to `dynPred`, which acts as the main workhorse function for `ROC`. Takes default arguments of `dynPred` if not supplied.
`progress`	should a progress bar be shown, showing the current progress of the ROC function ( to `progress = TRUE`.
`boot`	logical. Not currently used, legacy argument.

Value

A list of class ROC.joint consisting of:

Tstart: numeric denoting the start of the time window of interest; all dynamic predictions generated used longitudinal information up-to time T_{\mathrm{start}}.
delta: scalar which denotes length of interval to check, such that the window is defined by [T_{\mathrm{start}}, T_{\mathrm{start}}, + \delta].
candidate.u: candidate vector of failure times to calculate dynamic probability of surviving for each subject alive in data at time T_{\mathrm{start}}.
window.failures: numeric denoting the number of observed failures in [T_{\mathrm{start}}, T_{\mathrm{start}}, + \delta].
Tstart.alive: numeric denoting the risk set at Tstart.
metrics: a data.frame containing probabilistic thresholds with: TP true positives; FN false negatives; FP false positives; TN true negatives; TPR true positive rate (sensitivity); FPR false positive rate (1-specificity); Acc accuracy; PPV positive predictive value (precision); NPV negative predictive value; F1s F1 score and J Youden's J statistic.
AUC: the area under the curve.
BrierScore: The Brier score.
PE: The predicted error (taking into account censoring), loss function: square.
MH.acceptance: Raw acceptance percentages for each subject sampled.
MH.acceptance.bar: mean acceptance of M-H scheme across all subjects.
simulation.info: list containing information about call to dynPred.

Author(s)

James Murray (j.murray7@ncl.ac.uk).

Examples


data(PBC)
PBC$serBilir <- log(PBC$serBilir)
long.formulas <- list(serBilir ~ drug * time + (1 + time|id))
surv.formula <- Surv(survtime, status) ~ drug
family <- list('gaussian')
fit <- joint(long.formulas, surv.formula, PBC, family)
(roc <- ROC(fit, PBC, Tstart = 8, delta = 2, control = list(nsim = 25)))
plot(roc)

gmvjoint documentation built on Oct. 6, 2024, 1:07 a.m.