On Sun, 27 Jan 2002, Roseli A. Leandro wrote:
>
>
> Dear all
>
> Could someone explain to why weights does not works in lm and glm
> in the example below?
> Thanks in advance, Roseli.
>
>
> function(model){
>
> www<-fitted(model)
> lm(formula(model),weights=www)
>
> }
>
> The message error is: Error in eval(expr,envir,enclos): object
> "www" not found.
>
Because it's looking in the environment of the model formula.
That's why it can find the other variables in the model even though you
didn't specify a dataframe to search.
You can use
function(model){
www<-fitted(model)
update(model, weights=www)
}
which does work (although it arguably shouldn't).
The handling of weights= in these modelling functions is a horrible wart
on the language. Although they look like ordinary arguments they have
very strange scoping rules.
-thomas
______________________________________________
[EMAIL PROTECTED] mailing list
https://www.stat.math.ethz.ch/mailman/listinfo/r-help
