A factor is a special character vector where the elements have pre-defined groups or ‘levels’. You can think of these as qualitative or categorical variables:

x <- c("yellow", "red", "red", "blue", "yellow", "blue")
class(x)

## [1] "character"

x_fact <- factor(x) # factor() is a function
class(x_fact)

## [1] "factor"

Factors have levels (character types do not).

x

## [1] "yellow" "red"    "red"    "blue"   "yellow" "blue"

x_fact

## [1] yellow red    red    blue   yellow blue  
## Levels: blue red yellow

Note that levels are, by default, in alphanumerical order.

Extract the levels of a factor vector using levels():

levels(x_fact)

## [1] "blue"   "red"    "yellow"

A package called forcats is really helpful for working with factors.

factor() is from base R and as_factor() is from forcats

Both can change a variable to be of class factor.

• factor() will order alphabetically unless told otherwise.
• as_factor() will order by first appearance unless told otherwise.

If you are assigning your levels manually either function is fine!

x <- c("yellow", "red", "red", "blue", "yellow", "blue")
x_fact_2 <- as_factor(x)
x_fact_2

## [1] yellow red    red    blue   yellow blue  
## Levels: yellow red blue

# Compare to factor() method:
x_fact

## [1] yellow red    red    blue   yellow blue  
## Levels: blue red yellow

Aggregate (sum) across ethnicity and gender:

dropouts <-
  dropouts %>%
  group_by(CDS_CODE) %>%
  summarize(
    Freshman = sum(D9),
    Sophomore = sum(D10),
    Junior = sum(D11),
    Senior = sum(D12)
  )
dropouts

## # A tibble: 5,507 × 5
##    CDS_CODE       Freshman Sophomore Junior Senior
##    <chr>             <dbl>     <dbl>  <dbl>  <dbl>
##  1 01100170112607        1         1      0      1
##  2 01100170123968        0         0      0      0
##  3 01100170130401        3         5     12     24
##  4 01100170130419        1         2     13     36
##  5 01100170131581        0         0      0      0
##  6 01100176002000        0         0      0      0
##  7 01316090131755        0         0      0      0
##  8 01316170131763        0         0      0      2
##  9 01611190000001        0         1      0      1
## 10 01611190106401        0         0      0      0
## # ℹ 5,497 more rows

Pivot to long format:

dropouts <-
  dropouts %>%
  pivot_longer(
    !CDS_CODE,
    names_to = "grade",
    values_to = "n_dropouts"
  )
dropouts

## # A tibble: 22,028 × 3
##    CDS_CODE       grade     n_dropouts
##    <chr>          <chr>          <dbl>
##  1 01100170112607 Freshman           1
##  2 01100170112607 Sophomore          1
##  3 01100170112607 Junior             0
##  4 01100170112607 Senior             1
##  5 01100170123968 Freshman           0
##  6 01100170123968 Sophomore          0
##  7 01100170123968 Junior             0
##  8 01100170123968 Senior             0
##  9 01100170130401 Freshman           3
## 10 01100170130401 Sophomore          5
## # ℹ 22,018 more rows

head(dropouts)

## # A tibble: 6 × 3
##   CDS_CODE       grade     n_dropouts
##   <chr>          <chr>          <dbl>
## 1 01100170112607 Freshman           1
## 2 01100170112607 Sophomore          1
## 3 01100170112607 Junior             0
## 4 01100170112607 Senior             1
## 5 01100170123968 Freshman           0
## 6 01100170123968 Sophomore          0

Notice that grade is a chr variable. This indicates that the values are character strings.

R does not realize that there is any order related to the grade values. It will assume that it is alphabetical.

However, we know that the order is: freshman, sophomore, junior, senior.

Use set.seed() to take the same random sample each time.

set.seed(123)
dropouts_subset <- slice_sample(dropouts, n = 32)

Let’s make a plot first.

dropouts_subset %>%
  ggplot(mapping = aes(x = grade, y = n_dropouts)) +
  geom_boxplot() +
  theme_bw(base_size = 16) # make all labels size 16

OK this is very useful, but it is a bit difficult to read. We expect the values to be plotted by the order that we know, not by alphabetical order.

Currently grade is class character but let’s change that to class factor which allows us to specify the levels or order of the values.

dropouts_fct <-
  dropouts_subset %>%
  mutate(grade = factor(grade,
    levels = c("Freshman", "Sophomore", "Junior", "Senior")
  ))

dropouts_fct %>%
  pull(grade) %>%
  levels()

## [1] "Freshman"  "Sophomore" "Junior"    "Senior"

head(dropouts_fct)

## # A tibble: 6 × 3
##   CDS_CODE       grade     n_dropouts
##   <chr>          <fct>          <dbl>
## 1 45699716050231 Junior             0
## 2 45700036050330 Junior             0
## 3 12630401230069 Sophomore          0
## 4 09100900930131 Sophomore          1
## 5 15633216009179 Junior             0
## 6 33670330113647 Sophomore          0

Now let’s make our plot again:

dropouts_fct %>%
  ggplot(mapping = aes(x = grade, y = n_dropouts)) +
  geom_boxplot() +
  theme_bw(base_size = 16)

Now that’s more like it! Notice how the data is automatically plotted in the order we would like.

Tables grouped by a character are arranged alphabetically.

dropouts_subset %>%
  group_by(grade) %>%
  summarize(total_dropouts = sum(n_dropouts))

## # A tibble: 4 × 2
##   grade     total_dropouts
##   <chr>              <dbl>
## 1 Freshman               3
## 2 Junior                 2
## 3 Senior                13
## 4 Sophomore             12

Tables grouped by a factor are arranged by level.

dropouts_fct %>%
  group_by(grade) %>%
  summarize(total_dropouts = sum(n_dropouts))

## # A tibble: 4 × 2
##   grade     total_dropouts
##   <fct>              <dbl>
## 1 Freshman               3
## 2 Sophomore             12
## 3 Junior                 2
## 4 Senior                13

dropouts_fct %>% pull(grade) %>% levels() # By year order

## [1] "Freshman"  "Sophomore" "Junior"    "Senior"

dropouts_fct <- dropouts_fct %>%
  mutate(
    grade = fct_reorder(grade, n_dropouts, mean)
  )
dropouts_fct %>% pull(grade) %>% levels() # by increasing mean dropouts

## [1] "Junior"    "Freshman"  "Sophomore" "Senior"

The fct_count() function of the forcats package is helpful for checking that the proportions of each level for a factor are similar. Need the prop = TRUE argument otherwise just counts are reported.

dropouts_fct %>%
  pull(grade) %>%
  fct_count(prop = TRUE)

## # A tibble: 4 × 3
##   f             n     p
##   <fct>     <int> <dbl>
## 1 Junior       12 0.375
## 2 Freshman      7 0.219
## 3 Sophomore     7 0.219
## 4 Senior        6 0.188

• the factor class allows us to have a different order from alphanumeric for categorical data
• we can change data to be a factor variable using mutate and a factor creating function like factor() or as_factor
• the as_factor() is from the forcats package (first appearance order by default)
• the factor() base R function (alphabetical order by default)
• with factor() we can specify the levels with the levels argument if we want a specific order
• the fct_reorder({variable_to_reorder}, {variable_to_order_by}, {summary function}) helps us reorder a variable by the values of another variable
• arranging, tabulating, and plotting the data will reflect the new order

🏠 Class Website
💻 Lab

Image by Gerd Altmann from Pixabay