R - Moving from an hardcoded script to a function

Question

Let's imagine I have a data frame containing 2 types information (X# and Y#).

df = data.frame(matrix(rnorm(600), nrow=100))
colnames(df) <- c("X1", "X2", "Y1", "Y2", "Y3", "Y4")

I use two columns (below X1 and Y1) to group them in 9 categories (each column being splitted in 3 categories containing 1/3 or the rows) and store them in a new column cat11 (I deeply apologize for the poor code I show you, but I am a just a beginner in R).

df$tmpx <- cut2(df$X1, g=3)
levels(df$tmpx) <- c(1,2,3)
df$tmpy <- cut2(df$Y1, g=3)
levels(df$tmpy) <- c(1,2,3)

enum <- 1
for (x in sort(unique(df$tmpx)))
{
  for (y in sort(unique(df$tmpy)))
  {
    print(enum)
    df$cat11[df$tmpx == x & df$tmpy == y] <- enum
    enum <- enum + 1
  }
}

What I am struggling to do now is to run this code for a selection of other combinations (e.g X1,Y4 > cat14; X2,Y1 > cat21; X2,Y3 > cat23).

I have been trying using function as well as lapply, but unsuccessfully yet. I think I am missing something obvious.

Any help would be much appreciated.

Roland · Accepted Answer · 2016-08-19 10:10:17Z

1

First I create all combinations of X and Y columns:

combs <- expand.grid(names(df)[grep("X", names(df))],
                     names(df)[grep("Y", names(df))],
                     stringsAsFactors = FALSE)
#  Var1 Var2
#1   X1   Y1
#2   X2   Y1
#3   X1   Y2
#4   X2   Y2
#5   X1   Y3
#6   X2   Y3
#7   X1   Y4
#8   X2   Y4

Then I write a vectorized alternative to your approach and wrap it in a function:

library(Hmisc)
fun <- function(DF, col1, col2) {
  tmpx <- cut2(df[[col1]], g=3)
  tmpx <- as.integer(tmpx)

  tmpy <- cut2(df[[col2]], g=3)
  tmpy <- as.integer(tmpy)

  (tmpx - 1) * 3 + tmpy #some simple maths
}

Note how I use [[ to extract columns given as character strings programmatically. You can't use $ for this (this is a FAQ). Study help("[").

Then I use mapply to apply the function to all combinations:

df[, paste0("cat", 
            gsub("[[:alpha:]]*", "", combs[,1]),
            gsub("[[:alpha:]]*", "", combs[,2]))] <- mapply(fun, combs[,1], combs[,2], 
                                                             MoreArgs = list(DF = df))

mapply loops over all elements of its arguments and applies a function to them. E.g., the function is applied to X1/Y1, X2/Y1, ...

The most complicated part is creating the column names. I use a simple regular expression here and just remove all letters from the column names given in combs.

edited Aug 19, 2016 at 10:10

answered Aug 19, 2016 at 8:31

Roland

134k12 gold badges203 silver badges305 bronze badges

Sign up to request clarification or add additional context in comments.

4 Comments

thepule Over a year ago

This is a very elegant solution. A quick question: inside fun, would there be any difference in using df[,col1] instead of df[[col1]] to extract columns programmatically?

Roland Over a year ago

No, both would work (I think the latter is slightly more efficient).

user3541159 Over a year ago

Thanks Roland, very well done and explained. I still have a question. Why did you placed the "library(Hmisc)" within the function ?

Roland Over a year ago

@user3541159 You can put it outside the function.

Collectives™ on Stack Overflow

R - Moving from an hardcoded script to a function

1 Answer 1

4 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

4 Comments

Your Answer

Sign up or log in

Post as a guest

Related