R help - Oct 2004 - Weighted regresion using lm

Hi:

Could anyone help me to clarify this: are the weights normalized inside lm 
function (package:stats) before applied to the error term?  For example:
>lm (cost ~ material, weights=quatity, data=receipt)
will lm normalize quatity such that sum(quatity) = 1? I traced to lm.wfit and 
then the weights get transferred into a precompiled FORTRAN module so I
can't
figure out. Thanks!


> version
         _                   
platform sparc-sun-solaris2.9
arch     sparc               
os       solaris2.9          
system   sparc, solaris2.9   
status                       
major    1                   
minor    9.0                 
year     2004                
month    04                  
day      12                  
language R                


Yu Shao
Wadsworth Center
NY Department of Health
Albany, NY 12208

Yu Shao wrote:
> Hi:
> 
> Could anyone help me to clarify this: are the weights normalized inside lm 
> function (package:stats) before applied to the error term?  For example:
> 
> 
>>lm (cost ~ material, weights=quatity, data=receipt)
> 
> 
> will lm normalize quatity such that sum(quatity) = 1? I traced to lm.wfit
and
> then the weights get transferred into a precompiled FORTRAN module so I
can't
> figure out. Thanks!
> 

Did you look at ?lm.wfit?

        w: vector of weights (length 'n') to be used in the fitting
           process for the 'wfit' functions.  Weighted least squares is
           used with weights 'w', i.e., 'sum(w * e^2)' is
minimized.

So, no, the weights are not normalized.

--sundar
> 
> 
> 
>>version
> 
>          _                   
> platform sparc-sun-solaris2.9
> arch     sparc               
> os       solaris2.9          
> system   sparc, solaris2.9   
> status                       
> major    1                   
> minor    9.0                 
> year     2004                
> month    04                  
> day      12                  
> language R                
> 
> 
> Yu Shao
> Wadsworth Center
> NY Department of Health
> Albany, NY 12208
> 
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide!
http://www.R-project.org/posting-guide.html

On Thu, 28 Oct 2004, Yu Shao wrote:
> Could anyone help me to clarify this: are the weights normalized inside lm 
> function (package:stats) before applied to the error term?  For example:
> 
> >lm (cost ~ material, weights=quatity, data=receipt)
> 
> will lm normalize quatity such that sum(quatity) = 1? I traced to lm.wfit
and
> then the weights get transferred into a precompiled FORTRAN module so I
can't
> figure out. Thanks!
Yes you can, as you have the sources and the documentation.  (R is Open 
Source.)  You could also do a simple experiment like that below.

Weighted regression is not defined to use normalized weights, and this is
not done in R.  Note, though, that the fit and the coefficients are the
same whether you normalize or not.

y <- rnorm(100)
x <- runif(100)
w <- 1:100
w1 <- w/sum(w)
(fm <- lm(y ~x, weights=w))
(fm1 <- lm(y ~x, weights=w1))
summary(fm)
summary(fm1)

show that only the residual se and the residuals are affected.


-- 
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