Is this R bug real?

Question

I am quite confused... and thinking is this bug real? How can it be? I just want to make a vector of 0's and 1's.

Here's the source and the outcome

n.subj=1000
prop.aber = 0.9
n.measure = 3
n.subj.norm = n.subj*(1-prop.aber)
n.subj.aber = n.subj*prop.aber
labE <- rnorm(n.subj*n.measure, 0, 1)
labT_ <- c(rep(0, n.subj.norm), rep(1, n.subj.aber)); length(labT_)
sum(labT_ == 0); sum(labT_ == 1)


[1] 99
[1] 900

My common sense tells me that it should be 100 and 900!!!!!!?!?!?????

See `help("sample")` for what you might actually want to achieve. There is no bug here. — Roland, Mar 16 '16 at 09:04

score 1 · Accepted Answer · edited May 23 '17 at 12:07

1

Due to the floating point issues, n.subj.norm is not exactly 100. See this post for more information

See the below for an example:

n.subj.norm == 100
# FALSE
length(rep(0, 100))
# 100
length(rep(0, n.subj.norm))
# 99
length(rep(0, round(n.subj.norm, 0)))
# 100

edited May 23 '17 at 12:07

Community

1
1

answered Mar 16 '16 at 09:06

Raad

2,675
1
13
26

1

Also: `class(n.subj.norm); as.integer(n.subj.norm)` – Roland Mar 16 '16 at 09:09
Quite shocking. I never figured that would cause any problem in functions like rep beside making comparisons... – KH Kim Mar 16 '16 at 09:15

Is this R bug real?

1 Answers1