R help - Jan 2003 - Testing ``<=" in R

Hello,

I've encountered the following:
> n_500
> tau_.95
> (n*(1-tau))
[1] 25
> (n*(1-tau))<=25
[1] FALSE
> (n*(1-tau))==25
[1] FALSE

I'm using UNIX R Version 1.4.0, and also tested in out in Windows 1.6.0. Is 
this a bug?

Masha

On Fri, 31 Jan 2003, Maria N Kocherginsky wrote:
> Hello,
>
> I've encountered the following:
>
> > n_500
> > tau_.95
> > (n*(1-tau))
> [1] 25
> > (n*(1-tau))<=25
> [1] FALSE
> > (n*(1-tau))==25
> [1] FALSE
>
> I'm using UNIX R Version 1.4.0, and also tested in out in Windows
1.6.0. Is
> this a bug?
>
No.

You can't get exact floating point on a computer.
> n <- 500
> tau <- 0.95
> n*(1-tau)-25
[1] 2.1316282072803006e-14

Floating point is only accurate to about 15 decimal digits (53 binary
digits). There are only a few circumstances where it makes sense to ask if
two floating point numbers are equal.

If you care about getting n*(1-tau)==25 then either
1/ you know the numbers are integers, in which case you can round them
2/ you care about the difference between 25 and 25.00000000000002, in
which case you are in trouble


	-thomas

Maria N Kocherginsky <mkocherg at students.uiuc.edu> writes:
> Hello,
> 
> I've encountered the following:
> 
> > n_500
> > tau_.95
> > (n*(1-tau))
> [1] 25
> > (n*(1-tau))<=25
> [1] FALSE
> > (n*(1-tau))==25
> [1] FALSE
> 
> I'm using UNIX R Version 1.4.0, and also tested in out in Windows
1.6.0. Is
> this a bug?
No. Can 1/20 be represented exactly as a binary number?
>500*(1-tau)-25
[1] 2.131628e-14


-- 
   O__  ---- Peter Dalgaard             Blegdamsvej 3  
  c/ /'_ --- Dept. of Biostatistics     2200 Cph. N   
 (*) \(*) -- University of Copenhagen   Denmark      Ph: (+45) 35327918
~~~~~~~~~~ - (p.dalgaard at biostat.ku.dk)             FAX: (+45) 35327907