1

I'm not even sure how to give this a better, and not obviously duplicate type, title but I think this is a different question about expand.grid.

I have a list of variables for which I need a data.frame or list of every possible combination to feed into a bit of ordinal regression.

The list:

> indVars <- as.list(c("T.P","T.M","T.S","E"))

Desired output:

  > out List of (?)
: "T.P"
: "T.M"
: "T.S"
: "E"
: "T.P" "T.M"
: "T.P" "T.S"
: "T.P" "E"
.
.
.
: "T.P" "T.M" "T.S" "E"

Attempted:

  1. expand.grid(indVars) gives a single row

    > expand.grid(indVars) Var1 Var2 Var3 Var4 1 T.P T.M T.S E

  2. expand.grid(indVars,indVars) gives 16 rows of all two variable combinations but doesn't do 3 or four AND where indVars[i]==indVars[i] (so you get rows like

    > expand.grid(indVars,indVars)[1,] Var1 Var2 1 T.P T.P

  3. Logic says expand.grid(indVars,indVars,indVars,indVars) to give all up to combinations (256 of them) but again you end up with rows with multiple instances of the same indVar. For example:

    > expand.grid(indVars,indVars,indVars,indVars)[241,] Var1 Var2 Var3 Var4 241 T.P T.P E E

Request: Can someone point out how to expand this list of 4 variables into every combination of 1,2,3 and 4 of them with no duplicates?

BarneyC
  • 529
  • 4
  • 17
  • 1
    If your ultimate goal is to actually _fit_ models of all possible combinations of predictors, then have a look at `MuMIn::dredge`. `dredge` seems to be happy with ordinal regression objects (see ["List of supported models"](https://cran.r-project.org/web/packages/MuMIn/MuMIn.pdf)). See also [different possible combinations of variables for a generalized linear model](http://stackoverflow.com/questions/22664116/different-possible-combinations-of-variables-for-a-generalized-linear-model). – Henrik Aug 24 '16 at 09:11

1 Answers1

3

Possibly this is an XY problem and there is a better aproach to do the ordinal regression.

I suspect that order doesn't matter. Use combn:

res <- lapply(seq_along(indVars), combn, x = indVars, simplify = FALSE)
unlist(res, FALSE)

# [[1]]
# [1] "T.P"
# 
# [[2]]
# [1] "T.M"
# 
# [[3]]
# [1] "T.S"
# 
# [[4]]
# [1] "E"
# 
# [[5]]
# [1] "T.P" "T.M"
# 
# [[6]]
# [1] "T.P" "T.S"
# 
# [[7]]
# [1] "T.P" "E"  
# 
# [[8]]
# [1] "T.M" "T.S"
# 
# [[9]]
# [1] "T.M" "E"  
# 
# [[10]]
# [1] "T.S" "E"  
# 
# [[11]]
# [1] "T.P" "T.M" "T.S"
# 
# [[12]]
# [1] "T.P" "T.M" "E"  
# 
# [[13]]
# [1] "T.P" "T.S" "E"  
# 
# [[14]]
# [1] "T.M" "T.S" "E"  
# 
# [[15]]
# [1] "T.P" "T.M" "T.S" "E"
Roland
  • 127,288
  • 10
  • 191
  • 288