Why is this easy comparison false?

Question

Why does this simple statement evaluate to FALSE in R?

mean(c(7.18, 7.13)) == 7.155

Furthermore, what do I have to do in order to make this a TRUE statement? Thanks!

http://cran.r-project.org/doc/FAQ/R-FAQ.html#Why-doesn_0027t-R-think-these-numbers-are-equal_003f — Ben Bolker, May 10 '14 at 19:05
OK, @BenBolker's comments are somewhat more insightful than my "answer", but the idea is the same. — rbatt, May 10 '14 at 19:08
I'm sorry to be a little grumpy about this, but we shouldn't be answering this question again. Thanks to @T.C. for the link to the canonical duplicate ... — Ben Bolker, May 10 '14 at 19:09
@BenBolker I didn't search for the question before I answered it (maybe I should have), but tbh I'd never read an answer before (although I'm not surprised it already exists). — rbatt, May 10 '14 at 19:14
@BenBolker, that's a pretty unlikely Stack Overflow search term you shared there :-) — A5C1D2H2I1M1N2O1R2T1, May 12 '14 at 02:37

rbatt · Answer 1 · 2014-05-10T19:17:59.677

3

It's probably due to small rounding error. Rounding to the third decimal place shows that they are equal:

round(mean(c(7.18, 7.13)), 3) == 7.155

Generally, don't rely on numerical comparisons to give expected logical outputs :)

edited May 10 '14 at 19:17

answered May 10 '14 at 19:06

rbatt

score 3 · Accepted Answer · edited May 23 '17 at 11:57

Floating point arithmetic is not exact. The answer to this question has more information.

You can actually see this:

> print(mean(c(7.18,7.13)), digits=16)
[1] 7.154999999999999
> print(7.155, digits=16)
[1] 7.155

In general, do not compare floating point numbers for equality (this applies to pretty much every programming language, not just R).

You can use all.equal to do an inexact comparison:

> all.equal(mean(c(7.18,7.13)), 7.155)
[1] TRUE

2 Answers2