1

I am working in R would like some help on ranking with multiple criteria. Using the mtcars dataset - I want to generate a new column in this case based initially on the rank of mtcars$mpg, then in the case of ties for this to be decided by the rank of mtcars$qsec for example. I have mtcars["rank"] = NA then mtcars$rank=rank(mtcars$mpg) but not sure how to include how to deal with the ties. I've tried mtcars$rank=order(mtcars$mpg, mtcars$qsec) but not getting the outcome I want - I want the initial ranking for mtcars$mpg and in the event of ties for this to be decided by the lower ranking in mtcars$qsec. Thanks.

Emily
  • 93
  • 1
  • 6
  • 1
    Welcome to SO! Please provide a [minimal](http://stackoverflow.com/help/mcve) and [reproducible](http://stackoverflow.com/questions/5963269/how-to-make-a-great-r-reproducible-example) question, including sample data and code attempted. If you don't want to use your own data, consider a similar problem with a pre-installed dataset such as `mtcars` or `iris` (to name but two). – r2evans Jul 06 '16 at 18:29
  • Thanks - sorry I'm new to R and SO. Easier to use the mtcars dataset - I want to generate a new dataset or say column in this case based initially on the rank of `mtcars$mpg`, then in the case of ties for this to be decided by the rank of `mtcars$qsec` for example. I have `mtcars[rank] = NA` then `mtcars$rank=rank(mtcars$mpg)` but not sure how to include how to deal with the ties. – Emily Jul 06 '16 at 18:45
  • Additionally, `order(...)` allows multiple levels, so `order(mtcars$mpg, mtcars$qsec)` does what you (I?) think it should. – r2evans Jul 06 '16 at 18:50
  • 1
    I don't see in the help how to break the ties based on the value of another column? – Emily Jul 06 '16 at 18:55
  • @Emily please edit your question and add details provided in the comments. Thanks – ahajib Jul 06 '16 at 18:56
  • Thanks updated the question now. – Emily Jul 06 '16 at 19:27
  • Does `mtcars$mpgrnk[order(mtcars$mpg, mtcars$qsec)] <- 1:nrow(mtcars)` work? – r2evans Jul 06 '16 at 20:08
  • Yes, that seems to have done it! Just getting my head around differences between rank and order... Thanks! – Emily Jul 06 '16 at 20:32

1 Answers1

1

I would first order it based on mpg and qsec.

mtcars <- mtcars[order(mtcars$mpg, mtcars$qsec), ]

Ranking is now simply giving indexing to the dataframe.

mtcars$rank <- 1:nrow(mtcars)

AntonySamuelB
  • 143
  • 7