Re: [S] Residual Deviance and log-likelihood in survreg

Prof Brian Ripley (ripley@stats.ox.ac.uk)
Fri, 13 Feb 1998 13:53:46 GMT


> From: Paul Monaghan <pfm@scmp.scm.liv.ac.uk>
> Subject: [S] Residual Deviance and log-likelihood in survreg
>
> Does anybody know how the 'residual deviance' is calculated in the function survreg for fitting
> parametric survival models. It does not seem to be -2 log L!
>

It would not be for a glm fit, either. The definition of residual
deviance used here is

2[log L(saturated model) - log L(this model)]

(For a glm fit we also need to worry about scale factors: `deviance' is
certainly more complicated than -2 log L. Indeed, as a likelihood of
continuous data is only defined up to a multiplicative constant, the
the absolute value of -2 log L has no statistical meaning.)

I think that if you are seeing a `residual deviance' you have version 3.3
or earlier of S-PLUS. In that case the saturated model uses the same
parameters for the distribution shape, but with a mean for each case
exactly matching the observed value (for non-censored cases). This is
only useful if the distribution has a fixed shape, which is why 3.4 and 4.0
do not print a `residual deviance'.

Another example of why we ask posters to give the version of S-PLUS in
use!

Brian Ripley
-----------------------------------------------------------------------
This message was distributed by s-news@wubios.wustl.edu. To unsubscribe
send e-mail to s-news-request@wubios.wustl.edu with the BODY of the
message: unsubscribe s-news