\HeaderA{survreg.distributions}{Parametric Survival Distributions}{survreg.distributions}
\keyword{survival}{survreg.distributions}
\begin{Description}\relax
List of distributions for accelerated failure models. These are
location-scale families for some transformation of time. The entry
describes  the cdf \eqn{F}{} and density \eqn{f}{} of a canonical member of
the family.
\end{Description}
\begin{Usage}
\begin{verbatim}
survreg.distributions
\end{verbatim}
\end{Usage}
\begin{Format}\relax
There are three basic formats; only the first two are used in the
built-in distributions
\Tabular{ll}{
name:& name of distribution\\
variance:& Variance \\
init(x,weights,...):& Function returning an initial\\ mean and
& variance\\
deviance(y,scale,parms):& Function returning the deviance\\
density(x,parms):& Function returning \eqn{F}{},\\
& \eqn{1-F}{},\eqn{f}{},\eqn{f'/f}{},\eqn{f''/f}{}\\
quantile(p,parms):& Quantile function\\
scale:& Optional fixed value for scale parameter\\
}
and for transformations of the time variable
\Tabular{ll}{
name:& name of distribution\\
dist:& name of transformed distribution\\
trans:& transformation (eg log)\\
dtrans:& derivative of transformation\\
itrans:& inverse of transformation\\
scale:& Optional fixed value for scale parameter\\
}
For transformations of user-defined families use
\Tabular{ll}{
name:& name of distribution\\
dist:& transformed distribution in first format\\
trans:& transformation (eg log)\\
dtrans:& derivative of transformation\\
itrans:& inverse of transformation\\
scale:& Optional fixed value for scale parameter\\
}
\end{Format}
\begin{Details}\relax
There are four basic distributions:\code{extreme}, \code{gaussian},
\code{logistic} and \code{t}. The last three
are parametrised in the same way as the distributions already present in
\R. The extreme value cdf is
\deqn{F=1-e^{-e^t}.}{}

When the logarithm of survival time has one of the first three distributions
we obtain respectively \code{weibull}, \code{lognormal}, and
\code{loglogistic}. The Weibull distribution is not parameterised the
same way as in \code{\LinkA{rweibull}{rweibull}}.

The other predefined distributions are defined in terms of these. The
\code{exponential} and \code{rayleigh} distributions are Weibull
distributions with fixed \code{scale} of 1 and 0.5 respectively, and
\code{loggaussian} is a synonym for \code{lognormal}.

Parts of the built-in distributions are hardcoded in C, so the elements
of \code{survreg.distributions} in the first format above must not be
changed and new ones must not be added.  The examples show how to
specify user-defined distributions to \code{survreg}.
\end{Details}
\begin{SeeAlso}\relax
\code{\LinkA{survreg}{survreg}},
\code{\LinkA{pnorm}{pnorm}},\code{\LinkA{plogis}{plogis}}, \code{\LinkA{pt}{pt}}
\end{SeeAlso}
\begin{Examples}
\begin{ExampleCode}
## not a good fit, but a useful example
survreg(Surv(futime,fustat)~ecog.ps+rx,data=ovarian,dist='extreme')
## 
my.extreme<-survreg.distributions$extreme
my.extreme$name<-"Xtreme"
survreg(Surv(futime,fustat)~ecog.ps+rx,data=ovarian,dist=my.extreme)

## time transformation
survreg(Surv(futime,fustat)~ecog.ps+rx,data=ovarian,dist='weibull',scale=1)
my.weibull<-survreg.distributions$weibull
my.weibull$dist<-my.extreme
survreg(Surv(futime,fustat)~ecog.ps+rx,data=ovarian,dist=my.weibull,scale=1)

## change the transformation to work in years
## intercept changes by log(365), other coefficients stay the same
my.weibull$trans<-function(y) log(y/365)
my.weibull$itrans<-function(y) exp(365*y)
survreg(Surv(futime,fustat)~ecog.ps+rx,data=ovarian,dist=my.weibull,scale=1)

## Weibull parametrisation
y<-rweibull(1000, shape=2, scale=5)
survreg(Surv(y)~1, dist="weibull")
## survreg reports scale=1/2, intercept=log(5)
\end{ExampleCode}
\end{Examples}