R help - Jul 2006 - random section of samples based on group membership

Hi all,

I have a matrix of 474 rows (samples) with 565 columns (variables).
each of the 474 samples belong to one of 120 groups, with the
groupings as a column in the above matrix. For example, the group
column would be:

1
1
1
2
2
2
.
.
.
120
120

I  want to randomly select one from each group.  Not all the groups
have the same number of samples, some have 4, some 3 etc.  Is there a
function to do this, or would I need to write a looping statement to
look at each successive group?

I basically want to combine the randomly selected samples from the 120
groups into a new matrix in order to perform a cluster analysis.

Thanks,
Wade

Dear Wade,

Say that your groups are

groups <- sort(sample(1:10, 100, replace = TRUE))

Create a dummy

rows <- 1:length(groups)

Then

tapply( rows, groups, function(x) sample(x, 1))

does the trick to select the row numbers you need for your sampling.

Sincerely,

Carlos J. Gil Bellosta
http://www.datanalytics.com
http://www.data-mining-blog.com


Quoting Wade Wall <wade.wall at gmail.com>:
> Hi all,
>
> I have a matrix of 474 rows (samples) with 565 columns (variables).
> each of the 474 samples belong to one of 120 groups, with the
> groupings as a column in the above matrix. For example, the group
> column would be:
>
> 1
> 1
> 1
> 2
> 2
> 2
> .
> .
> .
> 120
> 120
>
> I  want to randomly select one from each group.  Not all the groups
> have the same number of samples, some have 4, some 3 etc.  Is there a
> function to do this, or would I need to write a looping statement to
> look at each successive group?
>
> I basically want to combine the randomly selected samples from the 120
> groups into a new matrix in order to perform a cluster analysis.
>
> Thanks,
> Wade
>
> ______________________________________________
> 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
> and provide commented, minimal, self-contained, reproducible code.
>

Well, how you do it might be a matter of taste with respect to how you
want the results.

You could try using "by" with "sample"

by(x,x[,3],function(y){y[sample(nrow(y),1),]})

This will return a list with one list element for each sample group.
You can the combine the list back into a matrix.

That's my naive solution; no doubt there will be half a dozen better
ways to go about it.

Also, some of the clustering functions I have seen will sample for you.


On 7/24/06, Wade Wall <wade.wall at gmail.com>
wrote:
> Hi all,
>
> I have a matrix of 474 rows (samples) with 565 columns (variables).
> each of the 474 samples belong to one of 120 groups, with the
> groupings as a column in the above matrix. For example, the group
> column would be:
>
> 1
> 1
> 1
> 2
> 2
> 2
> .
> .
> .
> 120
> 120
>
> I  want to randomly select one from each group.  Not all the groups
> have the same number of samples, some have 4, some 3 etc.  Is there a
> function to do this, or would I need to write a looping statement to
> look at each successive group?
>
> I basically want to combine the randomly selected samples from the 120
> groups into a new matrix in order to perform a cluster analysis.
>
> Thanks,
> Wade
>
> ______________________________________________
> 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
> and provide commented, minimal, self-contained, reproducible code.
>


-- 
Regards,

Mike Nielsen

On Mon, 24 Jul 2006 11:18:10 -0400,
"Wade Wall" <wade.wall at gmail.com> wrote:
> Hi all, I have a matrix of 474 rows (samples) with 565 columns
> (variables).  each of the 474 samples belong to one of 120 groups, with
> the groupings as a column in the above matrix. For example, the group
> column would be:
> 1 1 1 2 2 2 .  .  .  120 120
> I want to randomly select one from each group.  Not all the groups have
> the same number of samples, some have 4, some 3 etc.  Is there a
> function to do this, or would I need to write a looping statement to
> look at each successive group?
I use the following for that (some of it hacked from help("sample")):

".resample" <- function(x, size, ...) {
    if(length(x) <= 1) {
        if(!missing(size) && size == 0) x[FALSE] else x
    } else sample(x, size, ...)
}


"randpick" <- function(x, by, size = 1, ...)
{
    nx <- seq(nrow(x))
    ind <- unlist(tapply(nx, by, .resample, size, ...))
    x[nx %in% ind, ]
}


So, for instance:

R> randpick(Indometh, Indometh$Subject, 3)
   Subject time conc
2        1 0.50 0.94
7        1 3.00 0.12
11       1 8.00 0.05
15       2 1.00 0.70
16       2 1.25 0.64
19       2 4.00 0.20
25       3 0.75 1.16
29       3 3.00 0.22
32       3 6.00 0.08
34       4 0.25 1.85
43       4 6.00 0.07
44       4 8.00 0.07
48       5 1.00 0.39
54       5 6.00 0.10
55       5 8.00 0.06
58       6 0.75 1.03
64       6 5.00 0.13
65       6 6.00 0.10
R> randpick(Indometh, Indometh$Subject, 2)
   Subject time conc
8        1 4.00 0.11
10       1 6.00 0.07
14       2 0.75 0.71
20       2 5.00 0.25
23       3 0.25 2.72
28       3 2.00 0.39
39       4 2.00 0.40
43       4 6.00 0.07
48       5 1.00 0.39
52       5 4.00 0.11
57       6 0.50 1.44
66       6 8.00 0.09


The 'by' argument allows to sample within any combination of factors
desired.


Cheers,

-- 
Seb