NPbayesSurv function

Bayesian non-parametric estimation of a survival curve with right-censored data

Bayesian non-parametric estimation of a survival curve for right-censored data as proposed by Susarla and Van Ryzin (1976, 1978)

NPbayesSurv(time, censor,
  choice = c("exp", "weibull", "lnorm"), c = 1, parm,
  xlab = "Time", ylab = "Survival Probability", maintitle = "",
  cex.lab = 1.2, cex.axis = 1.0, cex.main = 1.5, cex.text = 1.2, lwd = 2)

Arguments

• time, censor: numeric vectors with (right-censored) survival times and 0/1 censoring indicators (1 for event, 0 for censored)

• choice: a character string indicating the initial guess ($S^*$) of the survival distribution

• c: parameter of the Dirichlet process prior

• parm: a numeric vector of parameters for the initial guess: rate parameter for exponential (see also Exponential), a two-element vector with shape

  and scale parameters for weibull (see also Weibull), a two-element vector with meanlog and sdlog parameters for log-normal (see also Lognormal). If not given, parameters for the initial guess are taken from the ML fit

• xlab, ylab: labels for axes of the plot

• maintitle: text for the main title

• cex.lab, cex.axis, cex.main, cex.text, lwd: graphical parameters

Returns

A vector corresponding to the parm argument

References

Susarla, V. and Van Ryzin, J. (1976). Nonparametric Bayesian estimation of survival curves from incomplete observations. Journal of the American Statistical Association, 71 (356), 897-902.

Susarla, V. and Van Ryzin, J. (1978). Large sample theory for a Bayesian nonparametric survival curve estimator based on censored samples. The Annals of Statistics, 6 (4), 755-768.

Author(s)

Emmanuel Lesaffre emmanuel.lesaffre@kuleuven.be , Arnošt Komárek arnost.komarek@mff.cuni.cz

Examples

## Nonparametric Bayesian estimation of a survival curve
## Homograft study, aortic homograft patients
data("graft", package = "icensBKL")

graft.AH <- subset(graft, Hgraft == "AH") # aortic homograft patients
time <- graft$timeFU[graft$Hgraft == "AH"]
censor <- graft$homo.failure[graft$Hgraft == "AH"]

  ## Initial guess: Weibull, c = 0.1 and 100
oldpar <- par(mfrow = c(1, 2))
NPbayesSurv(time, censor, "weibull", c = 100,
   xlab = "Follow-up time since the operation (years)", maintitle = "c = 100")
NPbayesSurv(time, censor, "weibull", c = 100,
   xlab = "Follow-up time since the operation (years)", maintitle = "c = 100")
par(oldpar)

  ## Initial guess: Exponential, c = 100
oldpar <- par(mfrow = c(1, 1))
NPbayesSurv(time, censor, "exp", c = 100,
   xlab = "Follow-up time since the operation (years)", maintitle = "Exp: c = 100")

  ## Initial guess: Log-normal, c = 100
NPbayesSurv(time, censor, "lnorm", c = 100,
   xlab = "Follow-up time since the operation (years)", maintitle = "Log-Normal: c = 100")
par(oldpar)