R help - May 2010 - How sample without replacement on more than one variables?

Hello All,

sample() only sample on one variable x. But I'm interested in sampling
more than one variable without replacement.

Suppose I have 3 vectors x, y, z. I want to draw samples from all
three vectors such that the combination of the three elements in each
draw is not the same as any previous draws. I could use expand.grid to
generate a vector out of the three vectors. But when the number of
vectors are large and the number of elements in some vectors are
large, it will be infeasible to do so.

If you know there is a method on sampling on more than one variables,
would you please let me know? Thank you!

-- 
Tom

thmsfuller066 at gmail.com wrote:
> Hello All,
> 
> sample() only sample on one variable x. But I'm interested in sampling
> more than one variable without replacement.
> 
> Suppose I have 3 vectors x, y, z. I want to draw samples from all
> three vectors such that the combination of the three elements in each
> draw is not the same as any previous draws. I could use expand.grid to
> generate a vector out of the three vectors. But when the number of
> vectors are large and the number of elements in some vectors are
> large, it will be infeasible to do so.
> 
> If you know there is a method on sampling on more than one variables,
> would you please let me know? Thank you!
> 
Can you give a reproducible example?  Since you suggested the method that is 
most reasonable, but it will not work in large cases, I suppose you'll have
to
draw independently from each vector one at a time, then somehow concatenate the 
results, perhaps as a character vector, even if the vectors are, say, integers. 
  Then repeat this process checking each time if your new vector is %in% the 
vector.

There may be a much better way, too, see if anyone else responds.

Also, you'll have to think about what a unique sample is.

If

x <- 1:3

y <- 2:4 ,

is x = 2, y = 3 the same as x = 3, y = 2?

Good luck,
Erik

This might help, depending on your exact needs:
> v1 <- sample(letters[1:2], 10, replace=TRUE)
> v2 <- sample(letters[3:4], 10, replace=TRUE)
> v3 <- sample(letters[5:6], 10, replace=TRUE)
> aa <- data.frame(v1=v1, v2=v2, v3=v3)
> aa
   v1 v2 v3
1   a  d  e
2   a  d  e
3   a  c  e
4   b  d  e
5   b  d  f
6   a  c  f
7   a  c  f
8   a  c  f
9   a  c  e
10  b  c  e
> bb <- unique(aa)
> bb
   v1 v2 v3
1   a  d  e
3   a  c  e
4   b  d  e
5   b  d  f
6   a  c  f
10  b  c  e

You can sample from the "bb" dataframe, or from the corresponding rows
of
the "aa" dataframe that are unique (1, 3, 4, 5, 6 and 10) which can be
obtained via rownames(bb).

Hth,
Adrian
-- 
View this message in context:
http://r.789695.n4.nabble.com/How-sample-without-replacement-on-more-than-one-variables-tp2227665p2227683.html
Sent from the R help mailing list archive at Nabble.com.

On Sun, 2010-05-23 at 00:56 -0700, dusadrian wrote:
> This might help, depending on your exact needs:
> > v1 <- sample(letters[1:2], 10, replace=TRUE)
> > v2 <- sample(letters[3:4], 10, replace=TRUE)
> > v3 <- sample(letters[5:6], 10, replace=TRUE)
> > aa <- data.frame(v1=v1, v2=v2, v3=v3)
And now is simple, sample the row of data frame
aa[sample(1:nrows(aa),3),]


-- 
Bernardo Rangel Tura, M.D,MPH,Ph.D
National Institute of Cardiology
Brazil