R help - Oct 2005 - Weights in survReg

Dear R users,

I am trying to find out what the function survReg does exactly with the 
Weights parameters. I looked in Terry Therneau documentations and other 
places and couldn't find anything. I tried to make an analogy with weighted 
OLS and assumed that the scale parameter Sigma in the accelerated 
failure-time model log(Time)= X*betas +Sigma*E, is of the form Sigma(i) = 
Sigma/Weights(i).

Can anyone confirm that or point me in the direction where I can find out 
for sure?

Thank you,
Sasha

Sasha,

This is R, which has survreg with weights (no caps).  You can find out for 
sure from the source code, but the help file says these are `observation 
weights'. I take that to mean weight=2 says `I have two cases like this'
in forming the likelihood.  The C code says `case weights', the usual term 
for that meaning.


On Wed, 19 Oct 2005, Alexander Antonyuk wrote:
> Dear R users,
>
> I am trying to find out what the function survReg does exactly with the
> Weights parameters. I looked in Terry Therneau documentations and other
> places and couldn't find anything. I tried to make an analogy with
weighted
> OLS and assumed that the scale parameter Sigma in the accelerated
> failure-time model log(Time)= X*betas +Sigma*E, is of the form Sigma(i)
> Sigma/Weights(i).
>
> Can anyone confirm that or point me in the direction where I can find out
> for sure?
>
> Thank you,
> Sasha
-- 
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