[R] linear coefficient combination stderr?

[Abhijit Dasgupta]

look at the function "estimable" in the package gmodels. This does the 
t.stat that you mention below.

Otherwise, if you wanted to do this by hand, :

beta = m$coef
V = summary(m)$cov.unscaled
sigma = summary(m)$sigma

covBeta = V*sigma^2

Now if your contrast is a%*%beta, then its se is 
sqrt(t(a)%*%covBeta%*%a) and so the t.stat below is

t.stat = (a%*%beta)/sqrt(t(a)%*%covBeta%*%a)
or, in your particular case

t.stat = (beta[2]+beta[3])/sqrt(t(c(0,1,1))%*%covBeta%*%c(0,1,1))

Abhijit

Søren Højsgaard wrote:
> Try the esticon function in the doBy package.
> Regards
> Søren
>
> ________________________________
>
> Fra: r-help-bounces at stat.math.ethz.ch på vegne af ivo welch
> Sendt: on 14-02-2007 04:21
> Til: r-help
> Emne: [R] linear coefficient combination stderr?
>
>
>
> dear r-experts---I have scrounged around in google (searching r-help)
> for the really obvious, but failed.  could someone please point me to
> whatever function computes the standard error of a linear combination
> of the coefficients?
>
> m=lm( y ~ x1 + x2 + x3 );
> t.stat= (coef(m)[2]+coef(m)[3])) / mystery.function( m, x2 + x3 );
>
> obviously, this does not work, but hopefully explains my intent.
> quick pointer appreciated.
>
> regards,
>
> /iaw
>
> ______________________________________________
> 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
> and provide commented, minimal, self-contained, reproducible code.
>
> ______________________________________________
> 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
> and provide commented, minimal, self-contained, reproducible code.
>