survAccuracyMeasures (version 1.2)

survAM.estimate: Nonparametric and Semiparametric estimates of accuracy measures for a risk prediction marker from survival data

Description

This function estimates the AUC, TPR(c), FPR(c), PPV(c), and NPV(c) for for a specific timepoint and marker cutoff value c using semiparametric or nonparametric estimates. Standard errors, and confidence intervals are also computed. Either analytic or bootstrap standard errors can be computed.

Usage

survAM.estimate(time, event, marker, data,  predict.time,   marker.cutpoint = 'median',  estimation.method = "IPW",  ci.method = "logit.transformed", se.method = "bootstrap", bootstraps = 1000,  alpha=0.05)

Arguments

time

time to event variable

event

indicator for the status of event of interest. event = 0 for censored observations, and event = 1 for event of interest.

marker

marker variable of interest

data

data frame in which to look for input variables.

predict.time

numeric value of the timepoint of interest for which to estimate the risk measures

marker.cutpoint

numeric value indicating the value of the cutpoint 'c' at which to estimate 'FPR', 'TPR', 'NPV' or 'PPV'. default is 'median' which takes cutpoint as the marker median.

estimation.method

Either "IPW" for non-parametric IPW estimates (default) or "Cox" for semi-parametric estimates that use a Cox proportional hazards model.

ci.method

character string of either 'logit.transformed' (default) or 'standard' indicating whether normal approximated confidence intervals should be calculated using logistic transformed values or the standard method.

se.method

Method to calculate standard errors for estimates. Options are "bootstrap" (default) or "asymptotic". Asymptotic estimates are based on large sample calculations and will not hold in small samples. Please see referenced papers for more information.

bootstraps

if se.method = 'bootstrap', number of bootstrap replicates to use to estimate the SE.

alpha

alpha value for confidence intervals. (1-alpha)*100 is alpha = 0.05.

Value

a list with components

estimates

point estimates for risk measures

se

standard errors for estimates

CIbounds

bounds for (1-alpha)*100 confidence interval

model.fit

if ESTmethod = "SP", object of type 'coxph' from fitting the model coxph(Surv(time, event)~Y)

cutoff, CImethod, SEmethod, predict.time, alpha

function inputs

References

Liu D, Cai T, Zheng Y. Evaluating the predictive value of biomarkers with stratified case-cohort design. Biometrics 2012, 4: 1219-1227.

Pepe MS, Zheng Y, Jin Y. Evaluating the ROC performance of markers for future events. Lifetime Data Analysis. 2008, 14: 86-113.

Zheng Y, Cai T, Pepe MS, Levy, W. Time-dependent predictive values of prognostic biomarkers with failure time outcome. JASA 2008, 103: 362-368.

Examples

data(SimData)

#non-parametric estimates
tmp <- survAM.estimate(time =survTime, 
                       event = status, 
                       marker = Y,
                       data = SimData,
                       estimation.method = "IPW",
                       predict.time = 2, 
                       marker.cutpoint = 0, 
                       bootstraps = 50)
tmp
tmp$estimates

#semi-parametric estimates
tmp <- survAM.estimate(time =survTime, 
                       event = status, 
                       marker = Y,
                       data = SimData,
                       estimation.method = "Cox",
                       predict.time = 2, 
                       marker.cutpoint = 0, 
                       bootstraps = 50)
                       
#semi-parametric estimates with asymptotic standard errors
tmp <- survAM.estimate(time =survTime, 
                       event = status, 
                       marker = Y,
                       data = SimData,
                       estimation.method = "Cox", 
                       se.method = "asymptotic", 
                       predict.time = 2, 
                       marker.cutpoint = 0, 
                       bootstraps = 50)