[R] Extracting t-tests on coefficients in lm

[Prof Brian Ripley]

On Wed, 20 Jun 2007, David C. Howell wrote:

> I am writing a resampling program for multiple regression using lm(). I
> resample the data 10,000 times, each time extracting the regression
> coefficients. At present I extract the individual regression
> coefficients using
>
>  brg = lm(Newdv~Teach + Exam + Knowledge + Grade + Enroll)
>  bcoef[i,] = brg$coef
>
> This works fine.
>
> But now I want to extract the t tests on these coefficients. I cannot
> find how these coefficients are stored, if at all. When I try
>    attributes(brg)
> I do not find the t values as the attributes of the object. Typing
> summary(brg) will PRINT the coefficients, their standard errors, t, and
> the associated probability. I would like to type something like
>    tcoeff[i,] = brg$tvalue
> but, of course, tvalue doesn't exist.
>
> Is there a simple way to extract, or compute if necessary, these values?

coef(summary(brg)) gives you the table, so coef(summary(brg))[,3] gives 
you the t values (but they are not t-tests per se).

?summary.lm would have told you this.

> Thanks,
> Dave Howell

-- 
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