0

I have data set and want to subset it to several data sets using all possible combinations of specific factor, giving that each combination should contain 4 levels and appear once.

here is some codes to generate very simple example: 

  data<-cbind(rep(1:8,each=2),matrix(nrow=16, ncol=4,rnorm(160, mean = 0, sd = 1)))
  colnames(data)<-LETTERS[1:5]

> data
     "A"          B          C          D           E
 [1,] 1 -0.07929477 -1.2946058 -1.4072064  0.57159386
 [2,] 1  1.83963909 -1.1723990  1.1232986  0.39483666
 [3,] 2 -0.36423210  1.3240148  1.3274450 -0.96929628

 [14,] 7  1.46756745 -0.7885119 -0.4218986 -1.25255228
 [15,] 8 -0.42291051  0.2915121  0.4320183  1.37582031
 [16,] 8 -0.40031215  0.4627476 -0.4145012  0.28700559

"A" is factor with 8 levels, i want to choose all possible 4 combinations ( i.e 1 2 3 4, 1 2 3 5 etc) out of the 8 levels and use that combinations to split "data" to several data sets to be used in further analysis.

Johan
  • 74,508
  • 24
  • 191
  • 319
hema
  • 725
  • 1
  • 8
  • 20

1 Answers1

2

Here you go:

## Generate all combinations of 4 integers between 1 and 8
ii <- combn(1:8, 4, simplify=FALSE)

## Use those combinations to pick out desired rows in data
x <- lapply(ii, function(II) data[data[,"A"] %in% II, ])

## Check that it worked
x[[1]]
#      A         B           C          D          E
# [1,] 1 2.7963535 -1.01141834  0.9133376 -1.3128354
# [2,] 1 1.9346950  0.85907646 -0.2222619 -0.8143439
# [3,] 2 2.2966139 -2.43140014 -0.4276004  0.4425973
# [4,] 2 0.9046734 -0.30193977 -0.1641523  1.2068400
# [5,] 3 0.8836684  2.59911207 -0.4339402  0.8922918
# [6,] 3 0.9004662  0.31611677  0.9300422 -0.4947400
# [7,] 4 1.0590443 -0.70879715 -0.2357002  1.0907113
# [8,] 4 1.6175373 -0.02734472  0.9151199 -0.8994856

x[[70]]
#      A          B          C           D           E
# [1,] 5  1.2375211 -0.8635894 -0.32504939 -0.38956232
# [2,] 5  1.0631257  1.7598401 -0.36029628  1.34065065
# [3,] 6  0.4014502 -0.9167007 -0.37284132  0.90406595
# [4,] 6  1.3352802 -1.4181380  0.27940665 -0.73645846
# [5,] 7  0.3892974  1.8418089  0.39443361  0.10841747
# [6,] 7  0.2152083 -0.4404339 -1.72481747 -0.03888857
# [7,] 8 -1.8517170  0.3844379 -0.04383212  1.02553227
# [8,] 8 -0.6770360 -2.0134745  1.71437731 -0.49894527
Josh O'Brien
  • 159,210
  • 26
  • 366
  • 455
  • Josh: thank you so much --- one more request how can i apply function on each data set say " the mean" for example and summaries the and put all the results in one vector – hema Oct 05 '13 at 00:27
  • 1
    @hema, `combn` has a `FUN` argument. Look into that as a start. Then, look at `lapply` and friends.... – A5C1D2H2I1M1N2O1R2T1 Oct 05 '13 at 01:30