Unexpected behaviour of greater and less than

Question

Possible Duplicate:
Why are these numbers not equal?

Can anyone explain to me, why R answers FALSE in the following case?

(1-0.9)>=0.1

How can I get TRUE for comparisons of that kind?

The most frequently asked FAQ 7.32: http://stackoverflow.com/questions/9508518/why-are-these-numbers-not-equal — Dieter Menne, Nov 11 '12 at 11:19

Romain Francois · Answer 1 · 2012-11-11T11:33:23.590

5

This is the classic R FAQ 7.31. You need all.equal

edited Nov 11 '12 at 11:33

answered Nov 11 '12 at 11:17

Romain Francois

17,432
3
51
77

+1, however I added some details in my answer to make this question somewhat self contained. – Paul Hiemstra Nov 11 '12 at 11:25
3

7.31 surely? Floating point and off-by-one errors combined! – Spacedman Nov 11 '12 at 11:29
thanks. I've edited to make less a fool of myself – Romain Francois Nov 11 '12 at 11:37

score 5 · Answer 2 · answered Nov 11 '12 at 11:22

This has to do with floating point precision. There is in essence an infinite amount of floating points, do representing them in the computer can only be done discretely, and thus with limited precision. To take this limited precision into account, use all.equal to make the comparison. As @RomainFracois said, this is very frequently asked question in R.

score 3 · Accepted Answer · answered Nov 11 '12 at 14:01

3

You could create your own binary operators to do what you're after and store them in your .Rprofile:

`%>=%` <- function(x, y) all.equal(x, y) | x > y
`%<=%` <- function(x, y) all.equal(x, y) | x < y

c(1-.9)>=.1
c(1-.9)%>=% .1

answered Nov 11 '12 at 14:01

Tyler Rinker

108,132
65
322
519

+1, neat trick to do this kind of comparison. Maybe `~=` would be an even better name. – Paul Hiemstra Nov 11 '12 at 18:07

Unexpected behaviour of greater and less than

3 Answers3