--begin inclusion ---
I am a student (and very to new to R) working on a senior design project that
is attempting to determine the demand distributions for single copy
newspaper draws at individual sales outlet locations. Our sales data is
right-censored, because sell-outs constitute a majority of the data, and we
are also testing the relevance of including covariates (weather,
seasonality, economic condition, etc.). We are trying to do MLE
calculations to determine each demand distribution and then use those
distributions for demand in the Newsvendor model. Is there any package (or
combination of packages) to help? We've been looking at survival...
--- end inclusion ---
The survival package has been most strongly influenced by medical
applications, where most of the data is not censored. The industrial
reliability literature, however, is filled with data sets more like yours, where
sometimes >90% is censored. This leads to different ways of thinking about
the
data. You might want to look at Escobar and Meeker, Statistical Methods for
Reliability Data, Wiley 1998, chapter 12 "Prediction of Future Random
Quantities". I think that many of their graphs may be relevant to your
situation.
They have an Splus package called censorreg, I don't know about an R
distribution. However, nearly everything can be done with survreg (from the
survival package), using predicted values from the models and building your own
graph.
Terry Therneau