[R] after lm-fit: equality of two regression coefficients test

[Prof Brian Ripley]

On Mon, 30 Aug 2004, Christoph Lehmann wrote:

> Hi
> 
> Let's assume, we have a multiple linear regression, such as the one 
> using the Scottish hills data (MASS, data(hills)):
> 
> one dependent variable: time
> two independent var (metric): dist, climb
> 
> if I am interested, after (!) fitting a lm:
> 
> 	my. lm <- lm(time ~ dist + climb, data = hills)
> 
> in the equivalence (or non-equivalence) of the two predictors "dist" and 
> "climb":
> 
> 	H0: dist = climb

I think you intend to ask if the *coefficients* in the fit should be 
equal, which is nonsense in this example of course.

> Is there any function in R, which lets me calculate this, in just giving 
> the lm-object "my.lm" and e.g. a vector such as c(1, -1), 
> operationalizing the hypothesis H0: t(c(1, -1)) %*% c(dist, climb) = 0 ?

library(car)
?linear.hypothesis


-- 
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