With a simple vector like
x <- sample(letters[1:3], size=20, replace=T)
I would extract the most frequent letter with something like
y <- table(x)
print(names(y)[y==max(y)])
"b"
However, using the same technique over a multidimensional dataframe does not work:
set.seed(5)
x <- data.frame(c1=sample(letters[1:3], size=30, replace=T),
c2=sample(letters[4:5], size=30, replace=T),
c3=sample(letters[6:10], size=30, replace=T))
y <- table(x)
print(names(y)[y==max(y)])
NULL
How can I extract the levels of c1, c2, and c3 that have the highest value in the contingency table?
I know I could convert the table to a dataframe and find the row where the Freq column is highest, but given the number of dimensions & levels in my dataset, doing the conversion to a dataframe would not fit in my RAM memory.
Edit: So my expected output in the second case would be c, d, j, as in:
z <- data.frame(y)
z[z$Freq==max(z$Freq), 1:3]
c1 c2 c3
27 c d j
But note that I cannot use the data.frame call on my data due to RAM issues.
You can use which with arr.ind = TRUE:
mapply("[",
dimnames(y),
as.data.frame(which(y == max(y), arr.ind = TRUE)))
# c1 c2 c3
#"c" "d" "j"
mapply("[",
dimnames(y),
as.data.frame(which(y == min(y), arr.ind = TRUE)))
# c1 c2 c3
# [1,] "a" "d" "f"
# [2,] "b" "d" "g"
# [3,] "c" "d" "g"
# [4,] "b" "e" "g"
# [5,] "a" "d" "h"
# [6,] "b" "d" "h"
# [7,] "c" "d" "h"
# [8,] "c" "e" "h"
# [9,] "a" "e" "i"
#[10,] "b" "e" "i"
#[11,] "c" "e" "i"