On Aug 16, 2011, at 10:48 AM, #HE YAO FENG VINCENT# wrote:
Hi all,
May i know does R has packages or code to run "Bayesian Relative
Survival Analysis"? I have look through Bayesian Survival
Analysis(2001) by Joseph George Ibrahim<http://www.google.com/search?tbo=p&tbm=bks&q=inauthor:%22Joseph+George+Ibrahim%22
>, Ming-Hui Chen<http://www.google.com/search?tbo=p&tbm=bks&q=inauthor:%22Ming-Hui+Chen%22
>, Debajyoti Sinha<http://www.google.com/search?tbo=p&tbm=bks&q=inauthor:%22Debajyoti+Sinha%22
> and would like to try out bayesian relative survival analysis in R.
From http://cran.r-project.org/web/packages/available_packages_by_name.html
, i know that the package relsurv<http://cran.r-project.org/web/packages/relsurv/index.html
> is for Relative survival and the package splinesurv<http://cran.r-project.org/web/packages/splinesurv/index.html
> is for Nonparametric bayesian survival analysis.
(For your information, relative survival is the method of choice for
estimating patient survival using data collected by population-based
cancer registries although its utility is not restricted to studying
cancer( Dickman and Adami 2006; Dickman et al. 2004)., which is a
concept defined by Berkson (1942) and Berkson & Gage (1950).Two data
files are required in order to estimate relative survival; a file
containing individual-level data on the patients and a file
containing expected probabilities of survival for a comparable
general population.)
There might be ( rather there _are_) arguments against the assertion
that "relative survival" is the "method of choice" for estimating
outcome even from population cancer stats. It obscures effects rather
than properly defining them when the observed survivals are close to
1.0 . If survival from a particular cancer at time T is 0.98 while
survival in the population is 0.995, the relative survival estimate is
obscuring the three-fold increase in risk of death to time T.
Sometimes you do want that estimate available when talking to
patients, but I think there is danger in hiding the excess risk by
using an estimate of 0.985 as the relative survival, because it gives
a signal that there is no remaining problem.
Both the paper, Spatial variation in prostate cancer survival in the
Northern and Yorkshire region of England using Bayesian relative
survival smoothing<http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2600701/
> and A Bayesian geoadditive relative survival analysis of registry
data on breast cancer mortality. <http://epub.ub.uni-muenchen.de/1881/1/paper_515.pdf
> use Bayesian relative survival analysis. Hence, i was wondering if
this is possible in R.
I have seen implementations of relative survival in R (and read at
least some of the articles making the above claims) but since I
disagree with those claims, I continue to focus my attention on the
methods available in survival and rms packages.
This will do a search of functions and rhelp postings:
http://search.r-project.org/cgi-bin/namazu.cgi?query=%22relative+survival%22&max=100&result=normal&sort=score&idxname=functions&idxname=Rhelp08&idxname=Rhelp10&idxname=Rhelp02
Thanks a lot.
Your Sincerely,
Vincent He
[[alternative HTML version deleted]]
David Winsemius, MD
West Hartford, CT
______________________________________________
R-help@r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.
