There seems to be a resurgence of interests for the nonparametric Bayes estimator of survival curve with censored data.
I got several emails asking about the R package NPBayes.
Therefore I updated the package I wrote 10 years ago, it is now working with more recent version of R (2.15.0, and 3.0.2).
NPBayes_0.7.tar.gz (source code, do not need compile) (released 6/12/2014)
NPBayes_0.8.tar.gz (released 9/19/2014)
References:
Susarla,
V and Van Ryzin, J. (1976). Nonparametric Bayesian
estimation of survival curves from incomplete observations. JASA 71
897-902.
Zhou, M. (2004). Nonparametric Bayes estimator of
survival functions from doubly/interval censored data. Statistica
Sinica 14, 533-546.
Zhou, M. and Luan, J. (2005)
Nonparametric
Bayes Estimator of Survival Function for
Right-Censoring and Left-Truncation Data Note: The parameter for the Dirichlet process prior is a measure on R+, in the package
NPBayes we took it to be a measure generated
by
the function B exp( - theta t), i.e. for any interval [a, b) the
measure is B[ exp(-theta a) - exp(-theta b) ] ,
where B
and theta are two scalar parameters. This could be changed to any
other positive measure on the R+, with some code modifications.
The prior is a Dirichlet process prior. Not mix of Dirichlet process priors.
The package only computes the mean of the posterior, no credible regions..