I have difficulties understanding the logic behind reorder().

Suppose Var is defined as below:

Var <- factor(c(0.2, 0.1, -0.1))

order(Var)

Now if I want to reorder it to c(1, 2, 3) I would run the following code, which works perfectly fine.

Needed_Order <- c(1, 2, 3)

Var <- reorder(Var, Needed_Order)

order(Var)

But it does not work if I want to reorder Var to c(3, 1, 2)

Needed_Order <- c(3,1,2)

Var <- reorder(Var, Needed_Order)

order(Var)

I expect to get 3 1 2 as the output of order(var) but it returns 2 3 1.

I think @prosoitos already has a great answer. I just wanted to illustrate why the reorder function exists and how it's useful.

Ordering Groups in a Plot

Let's consider the classic iris dataset

> data(iris)

> head(iris)

  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

and suppose we want to plot the Sepal.Width values, comparing by Species

boxplot(Sepal.Width ~ Species, iris)

but the order here is by species name, whereas we think the plot would look nicer if we ordered by the mean sepal width of each species. And that's where reorder is a quick, powerful solution for this:

iris$Species <- reorder(iris$Species, iris$Sepal.Width, FUN=mean)

boxplot(Sepal.Width ~ Species, iris)

What happened here is that the iris$Sepal.Width values corresponding to each level in iris$Species had the function mean applied to them and the result was attached to the factor as the scores attribute:

> attr(iris$Species, 'scores')

    setosa versicolor  virginica

     3.428      2.770      2.974

These scores were then used to rank (in ascending order) the levels in the factor, and assign them that order:

> levels(iris$Species)

[1] "versicolor" "virginica"  "setosa"

Note that this doesn't change the order of any of the data in the data frame, but only the order of the codes used in the factor. The FUN argument makes the reorder function quite general, so that one could order by min or max or whatever function you'd like to compute on the grouped data.

Overall, I think the key thing is that the second argument in the reorder function, which was thought to be a desired order, is instead for inputting the weights or values associated to each entry in the factor.

The function lvls_reorder() from the tidyverse package forcats does what you want:

Var <- factor(c(0.2, 0.1, -0.1))

Needed_Order <- c(3, 1, 2)

Var <- forcats::lvls_reorder(Var, Needed_Order)

order(Var)

Result

[1] 3 1 2

Explanations

Let's use an example with distinct elements for the values, the levels and the order positions of the levels to make it easier to visualize what is going on:

f <- factor(c(a = "A", b = "B", c = "C"))

f

# a b c 

# A B C 

# Levels: A B C



order(f)

# [1] 1 2 3

Now, let's use stats::reorder():

reorder(f, c(3, 1, 2))

# a b c 

# A B C 

# attr(,"scores")

# A B C 

# 3 1 2 

# Levels: B C A

reorder() assigns the values 3 1 2 as "scores" attributes for the levels A B C and reorders those levels according to these scores: reordering the scores to 1 2 3 reorders the levels to B C A.

Since order() returns a permutation which rearranges the factor into (by default) an ascending order, we get:

order(reorder(f, c(3, 1, 2)))

# [1] 2 3 1

In comparison, forcats::lvls_reorder() simply reorders the levels by indexing them with the values 3 1 2 (what you were trying to do):

lvls_reorder(f, c(3, 1, 2))

# a b c

# A B C

# Levels: C A B

Which gives the order:

order(lvls_reorder(f, c(3, 1, 2)))

# [1] 3 1 2