num <- rep(1:4, times = 2)

# The following lines return values differently
factor(num)
factor(num, labels = c("a", "b", "c", "d"))
factor(num, levels = c("a", "b", "c", "d"))

Question: What is the difference between levels and labels of a factor variable?

---

The “levels” argument allows us to change the orders (or the baseline) of the factor. In this case, “1” is the baseline category by default because it’s the smallest number among the four ones. Still we can change that by factor(data, levels = 4:1), and now “4” is the baseline category.

Thus, if something exist in the data but don’t exist in levels, they will be unrecognizable and R returns NAs.

On the other hand, “labels” labels the factor with another set of character strings, in the same order as levels. It doesn’t matter if the labels are completely different with the original data, but one should be careful not to be confused after relabeling:

# 1 is "a"
factor(data, labels = c("a", "b", "c", "d"))
# 4 is "a"
factor(data, levels = 4:1, labels = c("a","b","c","d"))  

Some experienced users may know the levels function as well. Be cautious when you are trying to use it to change the level attribute of a factor object. You can assign a new vector so that the category names are revised, but it DOES NOT really change the “levels” order of a factor. In fact, what it does is merely change the “labels”. For instance, say you want to change the baseline among species among the iris dataset:

species <- iris$Species

# You should do:
factor(species, levels = c("versicolor", "setosa", "virginica"))
# or:
relevel(species, "versicolor")

# You should NOT do:
levels(species) <- c("versicolor", "setosa", "virginica")

The latter will only change the “names”, so now what is called “versicolor” is actually corresponded with original “setosa”.