I want to gain thousands of random sampling data by randomizing the presence-absence data. Meantime, one important limition is that the row and column sums must be fixed. For example, the data "tst" is following: site1 site2 site3 site4 site5 site6 site7 site8 1 0 0 0 1 1 0 0 0 1 1 1 0 1 0 1 1 0 0 0 1 0 1 0 0 0 0 1 0 1 0 1 1 0 1 0 0 0 0 0 0 1 0 1 1 1 1 1 1 0 0 0 0 0 0 0 0 0 0 1 0 1 0 1 sum(tst[1,]) = 3, sum(tst[,1])=4, and so on. When I randomize the data, the first row sums must equal to 3, and the first column sums must equal to 4. The rules need to be applied to each row and column. How to get the new random sampling data? I have no idea. Thanks. [[alternative HTML version deleted]]
> -----Original Message----- > From: r-help-bounces at stat.math.ethz.ch [mailto:r-help-bounces at stat.math.ethz.ch] > On Behalf Of Zhang Jian > Sent: Saturday, July 07, 2007 12:31 PM > To: r-help > Subject: [R] random sampling with some limitive conditions? > > I want to gain thousands of random sampling data by randomizing the > presence-absence data. Meantime, one important limition is that the row and > column sums must be fixed. For example, the data "tst" is following: > site1 site2 site3 site4 site5 site6 site7 site8 1 0 0 0 1 1 0 0 0 1 1 1 0 > 1 0 1 1 0 0 0 1 0 1 0 0 0 0 1 0 1 0 1 1 0 1 0 0 0 0 0 0 1 0 1 1 1 1 1 1 0 0 > 0 0 0 0 0 0 0 0 1 0 1 0 1 > > sum(tst[1,]) = 3, sum(tst[,1])=4, and so on. When I randomize the data, the > first row sums must equal to 3, and the first column sums must equal to 4. > The rules need to be applied to each row and column. > How to get the new random sampling data? I have no idea. > Thanks. >You could reorder your table by stepping through your table a column at a time, and for each column randomly deciding to swap the current column with a column that has the same column total. Repeat this process for each row, i.e. for each row, randomly choose a row with the same row total to swap with. Here is some example code which is neither efficient nor general, but does demonstrate the basic idea. You will need to decide if this approach meets you needs. # I created a data file with your table (8x8) and read from it sites <- read.table("c:/R/R-examples/site_random_sample.txt", header=TRUE) sites # get row and column totals colsums <- apply(sites,2,sum) rowsums <- apply(sites,1,sum) # randomly swap columns for(i in 1:8) { if (runif(1) > .5) { swapcol<-sample(which(colsums==colsums[i]),1) temp<-sites[,swapcol] sites[,swapcol]<-sites[,i] sites[,i]<-temp } } # randomly swap rows for(i in 1:8) { if (runif(1) > .5) { swaprow<-sample(which(rowsums==rowsums[i]),1) temp<-sites[swaprow,] sites[swaprow,]<-sites[i,] sites[i,]<-temp } } sites Hope this is helpful, Dan Daniel Nordlund Bothell, WA USA
If I understand your problem, this might be a solution. Assign independent random numbers for row and column and use the corresponding ordering to assign the row and column indices. Thus row and column assignments are independent and the row and column totals are fixed. If cc and rr are respectively the desired row and column totals, with sum(cc)==sum(rr), then n = sum(cc) row.assign = rep(1:length(rr),rr)[order(runif(n))] col.assign = rep(1:length(cc),cc)[order(runif(n))] If you want many such sets of random assignments to be generated at once you can use a few more rep() calls in the expressions to generate multiple sets in the same way. (Do you actually want the assignments or just the tables?) Of course there are many other possible solutions since you have not fully specified the distribution you want. Alan Zaslavsky Harvard U> From: "Zhang Jian" <jzhang1982 at gmail.com> > Subject: [R] random sampling with some limitive conditions? > To: r-help <r-help at stat.math.ethz.ch> > > I want to gain thousands of random sampling data by randomizing the > presence-absence data. Meantime, one important limition is that the row and > column sums must be fixed. For example, the data "tst" is following: > site1 site2 site3 site4 site5 site6 site7 site8 1 0 0 0 1 1 0 0 0 1 1 1 0 > 1 0 1 1 0 0 0 1 0 1 0 0 0 0 1 0 1 0 1 1 0 1 0 0 0 0 0 0 1 0 1 1 1 1 1 1 0 0 > 0 0 0 0 0 0 0 0 1 0 1 0 1 > > sum(tst[1,]) = 3, sum(tst[,1])=4, and so on. When I randomize the data, the > first row sums must equal to 3, and the first column sums must equal to 4. > The rules need to be applied to each row and column. > How to get the new random sampling data? I have no idea. > Thanks.