risksetROC (version 1.0.4.1)

llCoxReg: Incident/Dynamic (I/D) ROC curve, AUC and integrated AUC (iAUC) estimation of censored survival data

Description

This function estimates the time-varying parameter estimate \(\beta(t)\) of non-proportional hazard model using local-linear Cox regression as discussed in Heagerty and Zheng, 2005.

Usage

llCoxReg(Stime, entry=NULL, status, marker, span=0.40, p=1, window="asymmetric")

Arguments

Stime

For right censored data, this is the follow up time. For left truncated data, this is the ending time for the interval.

entry

For left truncated data, this is the entry time of the interval. The default is set to NULL for right censored data.

status

Survival status.

marker

Marker value.

span

bandwidth parameter that controls the size of a local neighborhood.

p

1 if only the time-varying coefficient is of interest and 2 if the derivative of time-varying coefficient is also of interest, default is 1

window

Either of "asymmetric" or "symmetric", default is asymmetric.

Value

Returns a list of following items:

time

unique failure times

beta

estimate of time-varying parameter \(\beta(t)\) at each unique failure time.

Details

This function calculates the parameter estimate \(\beta(t)\) of non-proportional hazard model using local-linear Cox regression as discussed in Heagerty and Zheng, 2005. This estimation is based on a time-dependent Cox model (Cai and Sun, 2003). For p=1, the return item beta has two columns, the first column is the time-varying parameter estimate, while the second column is the derivative. However, if the derivative of the time-varying parameter is of interest, then we suggest to use p=2. In this case, beta has four columns, the first two columns are the same when p=1, while the last two columns estimates the coefficients of squared marker value and its derivative.

References

Heagerty, P.J., Zheng Y. (2005) Survival Model Predictive Accuracy and ROC curves Biometrics, 61, 92 -- 105

Examples

# NOT RUN {
data(pbc)
## considering only randomized patients
pbc1 <- pbc[1:312,]
## create new censoring variable combine 0,1 as 0, 2 as 1
survival.status <- ifelse( pbc1$status==2, 1, 0)
survival.time <- pbc1$fudays
pbc1$status1 <- survival.status
fit <- coxph( Surv(fudays,status1) ~ log(bili) +
                                     log(protime) +
                                     edema +
                                     albumin +
                                     age,
              data=pbc1 )
eta5 <- fit$linear.predictors
x <- eta5
nobs <- length(survival.time[survival.status==1])
span <- 1.0*(nobs^(-0.2))

# }
# NOT RUN {
bfnx1 <- llCoxReg(Stime=survival.time, status=survival.status, marker=x,
                   span=span, p=1)
plot(bfnx1$time, bfnx1$beta[,1], type="l", xlab="Time", ylab="beta(t)")
# }