[R] References verifying accuracy of R for basic statisticalcalculations and tests

[Prof Brian Ripley]

On Fri, 14 Jul 2006, Sean O'Riordain wrote:

> Please don't shoot!
> 
> q: would it be a good idea to use these datasets as a basis for some
> regression tests?

See package NISTnls

> On 14/07/06, Rau, Roland <Rau at demogr.mpg.de> wrote:
> > Hi,
> > > [mailto:r-help-bounces at stat.math.ethz.ch] On Behalf Of Corey Powell
> > >
> > > Do you know of any references that verify the accuracy of R
> > > for basic statistical calculations and tests.  The results of
> > > these studies should indicate that R results are the same as
> > > the results of other statistical packages to a certain number
> > > of decimal places on some benchmark calculations.
> >
> > I don't know of any references, but maybe you can somehow "verify the
> > accuracy of R" by running some analysis with the "NIST Statistical
> > Reference Datasets"; the URL is http://www.itl.nist.gov/div898/strd/
> > So maybe you can run the analyses mentioned there and say that R
> > (hopefully) returned the correct results.

`Correct' as in `as obtained by NIST'?  It is a considerable assumption 
that the reference results are 'correct' or 'accurate'.

I learnt from my work with analytical chemists that the outlying result 
could be the only reasonably accurate one: all the other analysts had made 
the same error.

> > Hope this helps,
> > Roland

-- 
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