[R] extractAIC.survreg problem with df

[Prof Brian Ripley]

A few more comments:

1) fit$residuals is a *component* of a list, not an attribute.

2) Since n is evaluated as zero, this only affects the constant in AIC
and model selection was unaffected.

3) fit$df is a vector, and you want sum(fit$df).

I've updated the code in R-patched.


On Wed, 21 Apr 2004, Prof Brian Ripley wrote:

> There has been a change in the survreg function since that was written 
> ....
> 
> On Wed, 21 Apr 2004, Stefano Calza wrote:
> 
> > Hi everybody.
> > 
> > I'm having problems with the extractAIC.survreg function and the edf
> > 
> > I get weird results which I think are due to the fact that the function defines edf as (from the stats package in 1.9.0 source code)
> > 
> > ...
> > ...
> > n <- length(fit$residuals)
> > edf <- n-fit$df.residual
> > ...
> > 
> > But in a survreg object there's no attribute residuals!!??
> > 
> > if in the function I use instead
> > 
> > ...
> > n <- length(residuals(fit)
> > ...
> > 
> > everything's fine, or better
> > 
> > ...
> > edf <- fit$df
> > ...
> > 
> > Am I missing something?


-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595