[R] precision of zero in R
Douglas Bates
bates at stat.wisc.edu
Sat Feb 3 16:59:15 CET 2007
On 2/3/07, pu chen <pchen at uni-bielefeld.de> wrote:
> hello,
> I have to calculate the determinants of near singular matrices. Presently R just stop at the precision of 2.16 e-16.
I'm not sure what you mean by that. Can you explain how you
determined that number, perhaps including some code?
It happens that on many systems that number is close to the relative
machine precision. For example, on my version of R
> .Machine$double.eps
[1] 2.220446e-16
> Can some one tell me how can I set the precision of zero in R ?
If you mean how could you change R so that it would give infinitely
precise floating point representations for which the relative machine
precision is zero I think the only answer is "by magic". In the real
world we need to work with the floating point representations on the
computers that we have access to. Perhaps you could rethink the
calculation you are attempting to perform.
More information about the R-help
mailing list