[R-lang] comparing regression coefficients
James S. Adelman
J.S.Adelman at warwick.ac.uk
Fri Sep 7 06:09:14 PDT 2007
On Fri, Sep 07, 2007 at 08:58:32PM +0800, Lngmyers wrote:
>
> All this talk about p-values for LME and mcmc reminds me of an old
> question. To compare the sizes of the coefficients within a single
> ordinary linear regression model, we can standardize them (by
> multiplying each by sd(x)/sd(y)) and look at the difference in their
> sizes. But we're not allowed to test whether this difference is
> statistically significant. I don't know enough math to know why not.
>
> Why couldn't we test the null hypothesis by resampling? Compute the
> standardized regression coefficients for each new sample, and count how
> many samples show a difference at least as large as the difference for
> the actual data.
>
> Is there any literature on this? Any a priori objections?
If I've understood your question correctly, you are asking about a linear
regression model with response, say z, and two predictors x and y:
K: z = a + mx + ny + error
and you wish to know whether H0: m=n. If so, anova(lm(z~x+y),lm(z~I(x+y)))
should be valid under the usual conditions.
--
James S. Adelman,
Department of Psychology,
University of Warwick,
COVENTRY, () ascii ribbon campaign - against html e-mail
CV4 7AL. /\ www.asciiribbon.org - against proprietary attachments
More information about the R-lang
mailing list