Skip to content

Problem in cluster sampling: 'mixed with negative subscripts'

9 messages · Chao Liu, Richard O'Keefe, Bert Gunter +3 more

#
Hi,

I was trying to do a cluster sampling but came across this error: Error in
xj[i] : only 0's may be mixed with negative subscripts. What is the cause
and how to get around? Thank you for your help!

Here is the code:

#simulate some data
y <- rnorm(20)
x <- rnorm(20)
z <- rep(1:5, 4)
w <- rep(1:4, each=5)
dd <- data.frame(id=z, cluster=w, x=x, y=y)
clusters <- split(dd, dd$cluster) #split into clusters
k <- length(clusters) #length of clusters
# This function generates a cluster sample
clsamp <- function() dd[unlist(clusters[sample.int(k, k, replace = TRUE)],
use.names = TRUE), ]
clsamp()


I got this error: Error in xj[i] : only 0's may be mixed with negative
subscripts.


Best,

Chao
#
1. Thanks for the example.
2. Good opportunity to learn (more) about debugging in R. See ?debug or
?browser()
3. Hint: what do you think ... unlist(clusters[...  gives?  (you are using
it as an index for subscripting dd)

Cheers,
Bert

Bert Gunter

"The trouble with having an open mind is that people keep coming along and
sticking things into it."
-- Opus (aka Berkeley Breathed in his "Bloom County" comic strip )
On Sat, Dec 19, 2020 at 10:58 AM Chao Liu <psychaoliu at gmail.com> wrote:

            

  
  
#
Hi Chao,
You have discovered one of the surprising things about the extraction
operator "[". It expects to get an object consisting of integers (like
1,2,3,...) or logical values (TRUE,FALSE or 0,1). As you have passed
the _values_ of your cluster, it can't deal with the negative numbers
as they don't index anything in the original object. If you get rid of
the negative values by applying the abs() function, it will seem to
work, but you aren't getting what you expected or anything sensible.
Here's an example:

# get a vector of positive and negative real numbers
x1<-rnorm(10)
x1
[1] -0.2174320 -1.3185389  0.4049751  0.4780766 -1.6317983  3.4265246
[7]  2.0721620  1.1590961  0.9896266  0.5672552
# try to index it with its values
x[x]
# Error in x[x] : only 0's may be mixed with negative subscripts
# now change the negative values to positive ones
 x1[abs(x1)]
# No error, but you only get some of the values!
[1] -0.2174320 -0.2174320  0.4049751 -1.3185389 -0.2174320
abs(x1)
[1] 0.2174320 1.3185389 0.4049751 0.4780766 1.6317983 3.4265246 2.0721620
[8] 1.1590961 0.9896266 0.5672552
# What the extraction operator does is attempt to get valid {positive}
integer indices
# or zeros. Then it can use the positive values and discard the zeros
as.integer(abs(x1))
[1] 0 1 0 0 1 3 2 1 0 0

Now the error message makes a lot more sense.

Jim
On Sun, Dec 20, 2020 at 5:58 AM Chao Liu <psychaoliu at gmail.com> wrote:
#
Jim and Bert,

Thank you for your help. It was an indexing problem. I was able to solve it
by changing the split line to: clusters <- split(seq_len(nrow(dd)),
dd$cluster).
Thank you.

Best,

Chao
On Sat, Dec 19, 2020 at 4:19 PM Jim Lemon <drjimlemon at gmail.com> wrote:

            

  
  
#
More accurately, in x[i] where x and i are simple vectors,
i may be a mix of positive integers and zeros
  where the zeros contribute nothing to the result
or it may be a MIX of negative integers and zeros
  where the zeros contribute nothing to the result
  and -k means "do not include element k".
It would be nice to write things like
  x[c(-1,1)]
meaning to copy everything except element 1, then
element 1.  But that is not allowed.
On Sun, 20 Dec 2020 at 10:20, Jim Lemon <drjimlemon at gmail.com> wrote:

            

  
  
#
Inline...
On Sat, Dec 19, 2020 at 9:22 PM Richard O'Keefe <raoknz at gmail.com> wrote:

            
But c(x[-1], x[1])  is, which is not so terrible, after all...

Cheers,
Bert

  
  
#
It does remind me of counting on one's fingers, though.

Jim
On Sun, Dec 20, 2020 at 4:38 PM Bert Gunter <bgunter.4567 at gmail.com> wrote:
But c(x[-1], x[1])  is, which is not so terrible, after all...
#
I've never mastered using negative indexes with my fingers, though...
On December 19, 2020 11:32:46 PM PST, Jim Lemon <drjimlemon at gmail.com> wrote:

  
    
#
On Sun, 20 Dec 2020 at 06:22, Richard O'Keefe <raoknz at gmail.com> wrote:

            
Yes, index 0 doesn't get an error or a warning. I think it should.
Eg.

(v <- 1:4)
[1] 1 2 3 4

v[c(0,2,4)] <- c(0,2)
v
[1] 1 0 3 2 # a surprise ?

Regards
Martin