I want to assign factor levels depending on the string weight. Normally R appears to assign factor levels alphabetically:
set.seed(54)
x <- sample(1:10, 5000, replace = TRUE)
x <- "levels<-"(as.factor(x), LETTERS[1:10])
> summary(x)
A B C D E F G H I J
524 508 519 489 477 496 507 526 473 481
I can reorder the factor levels and reassign them like this:
l <- data.frame(x=summary(x), old.levels=names(summary(x)),
row.names = NULL)
l <- transform(l[order(summary(x)), ],
new.levels=LETTERS[1:10])
levels(x) <- l[order(l$old.levels), 3]
> summary(x)
I G H D B E F J A C
524 508 519 489 477 496 507 526 473 481
But by this I haven't changed the factor values:
> summary(as.factor(as.numeric(x)))
1 2 3 4 5 6 7 8 9 10
524 508 519 489 477 496 507 526 473 481
How can I get smartly what I want?
Default behaviour for factor is indeed to order the levels and then assign
labels.
set.seed(54)
x <- sample(letters[1:10], 5000, replace = TRUE)
f1 <- factor(x, labels = LETTERS[1:10])
f2 <- factor(x, levels = sort(unique(x)), LETTERS[1:10])
summary(f1)
#> A B C D E F G H I J
#> 524 508 519 489 477 496 507 526 473 481
identical(f1, f2)
#> [1] TRUE
If you just want the labels assigned in frequency order, you can do that by reordering the labels when creating the factor:
f3 <- factor(x, levels = sort(unique(x)), LETTERS[1:10][order(table(x))])
summary(f3)
#> I E J D F G B C A H
#> 524 508 519 489 477 496 507 526 473 481
If you want the labels in frequency order and the levels sorted alphabetically, order the levels during factor creation instead:
f4 <- factor(x, levels = sort(unique(x))[order(table(x))], LETTERS[1:10])
summary(f4)
#> A B C D E F G H I J
#> 473 477 481 489 496 507 508 519 524 526
Created on 2018-03-16 by the reprex package (v0.2.0).