[R] Signs of loadings from princomp on Windows

[Prof Brian Ripley]

On 16 Sep 2004, Peter Dalgaard wrote:

> Tony Plate <tplate at blackmesacapital.com> writes:
> 
> > difference?"  (The fact that various people chimed in to say they
> > could reproduce the behavior that bothered you, but didn't bother dig
> > deeper suggests it didn't bother them that much, which further
> > suggests that you are the person most motivated by this and thus the
> > best candidate for investigating it further...)
> 
> It could be well worth finding though. It's most likely a reliance on
> uninitialized memory (the only other explanation I can think of is if
> something is tampering with the FPU control word). That may be mostly
> harmless if the algorithm converges - the sign is not well defined
> anyway - but we have recently fixed a couple of bugs where the
> occasional injection of NaN or Inf values caused real trouble.

valgrind found nothing, though, so the uninitialized memory is probably
not in R.

I suspect a problem with the version of msvcrt.dll in use (and probably
related to use of the full FPU precision in R).

And I do think the OP is making a very big deal out of something that is 
documented on the help page.

-- 
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