[R] decimal accuracy in pnorm( )

Prof Brian Ripley ripley at stats.ox.ac.uk
Thu Aug 10 08:07:44 CEST 2006


On Thu, 10 Aug 2006, Chronopoulos Dimitris wrote:

> Dear R users
> 
> Is there any way to increase the decimal accuracy for the normal 
> probability distribution? When one needs an accurate p-value for 
> instance this is provided by
> 
> pnorm(10,lower.tail=F)
> [1] 7.619853e-24
> 
> However, what happens when instead of a P[X<x], a more accurate P[X>=x] 
> is the objective.

That is P[X>=x] !

If you meant that 

> pnorm(10,lower.tail=TRUE)
[1] 1

then the problem is your computer, which has no representable numbers 
between 1-1e-16 and 1, and so correctly chose the nearest representable 
number.

A basic understanding of numerical methods is necessary to do statistical 
calculations accurately: perhaps your university offers courses in the 
area?

-- 
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



More information about the R-help mailing list