R Language Split function Basic usage of split
Example
split allows to divide a vector or a data.frame into buckets with regards to a factor/group variables. This ventilation into buckets takes the form of a list, that can then be used to apply group-wise computation (for loops or lapply/sapply).
First example shows the usage of split on a vector:
Consider following vector of letters:
testdata <- c("e", "o", "r", "g", "a", "y", "w", "q", "i", "s", "b", "v", "x", "h", "u")
Objective is to separate those letters into voyels and consonants, ie split it accordingly to letter type.
Let's first create a grouping vector:
vowels <- c('a','e','i','o','u','y')
letter_type <- ifelse(testdata %in% vowels, "vowels", "consonants")
Note that letter_type has the same length that our vector testdata.
Now we can split this test data in the two groups, vowels and consonants :
split(testdata, letter_type)
#$consonants
#[1] "r" "g" "w" "q" "s" "b" "v" "x" "h"
#$vowels
#[1] "e" "o" "a" "y" "i" "u"
Hence, the result is a list which names are coming from our grouping vector/factor letter_type.
split has also a method to deal with data.frames.
Consider for instance iris data:
data(iris)
By using split, one can create a list containing one data.frame per iris specie (variable: Species):
> liris <- split(iris, iris$Species)
> names(liris)
[1] "setosa" "versicolor" "virginica"
> head(liris$setosa)
Sepal.Length Sepal.Width Petal.Length Petal.Width Species
1 5.1 3.5 1.4 0.2 setosa
2 4.9 3.0 1.4 0.2 setosa
3 4.7 3.2 1.3 0.2 setosa
4 4.6 3.1 1.5 0.2 setosa
5 5.0 3.6 1.4 0.2 setosa
6 5.4 3.9 1.7 0.4 setosa
(contains only data for setosa group).
One example operation would be to compute correlation matrix per iris specie; one would then use lapply:
> (lcor <- lapply(liris, FUN=function(df) cor(df[,1:4])))
$setosa
Sepal.Length Sepal.Width Petal.Length Petal.Width
Sepal.Length 1.0000000 0.7425467 0.2671758 0.2780984
Sepal.Width 0.7425467 1.0000000 0.1777000 0.2327520
Petal.Length 0.2671758 0.1777000 1.0000000 0.3316300
Petal.Width 0.2780984 0.2327520 0.3316300 1.0000000
$versicolor
Sepal.Length Sepal.Width Petal.Length Petal.Width
Sepal.Length 1.0000000 0.5259107 0.7540490 0.5464611
Sepal.Width 0.5259107 1.0000000 0.5605221 0.6639987
Petal.Length 0.7540490 0.5605221 1.0000000 0.7866681
Petal.Width 0.5464611 0.6639987 0.7866681 1.0000000
$virginica
Sepal.Length Sepal.Width Petal.Length Petal.Width
Sepal.Length 1.0000000 0.4572278 0.8642247 0.2811077
Sepal.Width 0.4572278 1.0000000 0.4010446 0.5377280
Petal.Length 0.8642247 0.4010446 1.0000000 0.3221082
Petal.Width 0.2811077 0.5377280 0.3221082 1.0000000
Then we can retrieve per group the best pair of correlated variables: (correlation matrix is reshaped/melted, diagonal is filtered out and selecting best record is performed)
> library(reshape)
> (topcor <- lapply(lcor, FUN=function(cormat){
correlations <- melt(cormat,variable_name="correlatio);
filtered <- correlations[correlations$X1 != correlations$X2,];
filtered[which.max(filtered$correlation),]
}))
$setosa
X1 X2 correlation
2 Sepal.Width Sepal.Length 0.7425467
$versicolor
X1 X2 correlation
12 Petal.Width Petal.Length 0.7866681
$virginica
X1 X2 correlation
3 Petal.Length Sepal.Length 0.8642247
Note that one computations are performed on such groupwise level, one may be interested in stacking the results, which can be done with:
> (result <- do.call("rbind", topcor))
X1 X2 correlation
setosa Sepal.Width Sepal.Length 0.7425467
versicolor Petal.Width Petal.Length 0.7866681
virginica Petal.Length Sepal.Length 0.8642247
