R help - Nov 2004 - Problems with the sort function!?

Hello All.

I am running R 2.0.0 with a Win XP operating system. The same problem occured
with R 1.9.1 but not with R.1.3.1 on a Win NT computer.

My request is about the following problem with the sort function:

The vector price is a vector of 789 asset prices of the following form,
> mode(price)
[1] "numeric"
> price[1:10]
 [1] 1.1315 1.1314 1.1313 1.1314 1.1315 1.1316 1.1315 1.1317
 [9] 1.1316 1.1316

Then the returns are defined to be the differences,
> ret <- diff(price)
> ret[1:10]
 [1] -1e-04 -1e-04  1e-04  1e-04  1e-04 -1e-04  2e-04 -1e-04
 [9]  0e+00  0e+00

Here R recognizes the first two entries and also the second ones to be equal
(I did not found any two equal entries, where R didnt recognized this
relation),
> ret[1]==ret[2]
[1] TRUE
> ret[3]==ret[4]
[1] TRUE

But after sorting the vector, this ist not longer true,
> s <- sort(ret)
> s[1:10]
 [1] -5e-04 -5e-04 -5e-04 -5e-04 -5e-04 -4e-04 -4e-04 -4e-04
 [9] -4e-04 -4e-04
> s[1]==s[2]
[1] FALSE

Taking the difference of the first two entries yields
> s[1]-s[2]
[1] -2.220446e-16

which seems to be a numerical artefact.

Is this a bug, or just a handling error by me?

Thanks a lot for your answers,
Kay Frederik Pilz
Ruhr-Universitaet Bochum
Germany
Kay.Pilz at rub.de

> x <- 1.00000001
> all.equal(x, 1)
[1] TRUE
> x == 1
[1] FALSE

Reading help(all.equal) will tell you that the tolerance level by
default is around 1.490116e-08.


On Fri, 2004-11-05 at 13:48, Kay Pilz wrote:
> Hello All.
> 
> I am running R 2.0.0 with a Win XP operating system. The same problem
occured
> with R 1.9.1 but not with R.1.3.1 on a Win NT computer.
So, which version are you having this problem with ?
> My request is about the following problem with the sort function:
> 
> The vector price is a vector of 789 asset prices of the following form,
> 
> > mode(price)
> [1] "numeric"
> > price[1:10]
>  [1] 1.1315 1.1314 1.1313 1.1314 1.1315 1.1316 1.1315 1.1317
>  [9] 1.1316 1.1316
> 
> Then the returns are defined to be the differences,
> 
> > ret <- diff(price)
> > ret[1:10]
>  [1] -1e-04 -1e-04  1e-04  1e-04  1e-04 -1e-04  2e-04 -1e-04
>  [9]  0e+00  0e+00
> 
> Here R recognizes the first two entries and also the second ones to be
equal
> (I did not found any two equal entries, where R didnt recognized this
> relation),
> 
> > ret[1]==ret[2]
> [1] TRUE
> > ret[3]==ret[4]
> [1] TRUE
> 
> But after sorting the vector, this ist not longer true,
> 
> > s <- sort(ret)
> > s[1:10]
>  [1] -5e-04 -5e-04 -5e-04 -5e-04 -5e-04 -4e-04 -4e-04 -4e-04
>  [9] -4e-04 -4e-04
> > s[1]==s[2]
> [1] FALSE
> 
> Taking the difference of the first two entries yields
> 
> > s[1]-s[2]
> [1] -2.220446e-16
> 
> which seems to be a numerical artefact.
> 
> Is this a bug, or just a handling error by me?
> 
> Thanks a lot for your answers,
> Kay Frederik Pilz
> Ruhr-Universitaet Bochum
> Germany
> Kay.Pilz at rub.de
> 
> ______________________________________________
> R-help at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-help
> PLEASE do read the posting guide!
http://www.R-project.org/posting-guide.html
>

Kay Pilz <Kay.Pilz at ruhr-uni-bochum.de> writes:
> Hello All.
> 
> I am running R 2.0.0 with a Win XP operating system. The same problem
occured
> with R 1.9.1 but not with R.1.3.1 on a Win NT computer.
> 
> My request is about the following problem with the sort function:
...
> Taking the difference of the first two entries yields
> 
> > s[1]-s[2]
> [1] -2.220446e-16
> 
> which seems to be a numerical artefact.
> 
> Is this a bug, or just a handling error by me?
You, I think. Consider for instance

> x <- c(1.313,1.314,1.315,1.316)
> diff(x) - 1e-3
[1]  1.118897e-16 -1.101549e-16  1.118897e-16
> sort(diff(x) - 1e-3)
[1] -1.101549e-16  1.118897e-16  1.118897e-16
> diff(sort(diff(x) - 1e-3)[1:2])
[1] 2.220446e-16


-- 
   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