Mastering Categorical Data in R: A Guide to forcats

Handling Factors and Levels Systematically

What will we learn?

• Inspection: How to check factor levels and frequencies.
• Cleaning: Removing unused or empty levels.
• Reordering: Changing the order of levels for better visualization.
• Recoding: Renaming or collapsing multiple levels into meaningful groups.

---

1. Setup and Data Inspection

Factors are used to handle categorical variables with fixed and known sets of values. We will use the gss_cat dataset.

library(tidyverse)
theme_set(theme_minimal())

# View the General Social Survey dataset
# view(gss_cat)

# Check the levels of a factor variable
levels(gss_cat$race)

[1] "Other"          "Black"          "White"          "Not applicable"

# View frequency table
gss_cat %>% 
  select(race) %>% 
  table()

race
         Other          Black          White Not applicable 
          1959           3129          16395              0 

---

2. Removing Unused Levels

Sometimes factor levels exist in the metadata but have zero observations in the data. We can clean these up using fct_drop.

## remove unused levels  
gss_cat %>% 
  mutate(race = fct_drop(race)) %>% 
  select(race) %>% 
  table()

race
Other Black White 
 1959  3129 16395 

---

3. Modifying Factor Order

Changing the order of factors is essential for creating readable and professional plots.

Manual Reordering (fct_relevel)

Move specific levels to a custom position (e.g., bringing “White” and “Black” to the front).

gss_cat %>% 
  mutate(race = fct_drop(race)) %>% 
  mutate(race = fct_relevel(race, c("White", "Black", "Other"))) %>% 
  select(race) %>% 
  table()

race
White Black Other 
16395  3129  1959 

Order by Frequency (fct_infreq)

Automatically sort factors based on the number of observations.

## order the factors by frequency and reverse
gss_cat %>% 
  mutate(marital = fct_infreq(marital)) %>% 
  mutate(marital = fct_rev(marital)) %>% 
  ggplot(aes(marital)) +
  geom_bar(fill = "steelblue") +
  theme_classic(base_size = 20)

Order by Value in Another Variable (fct_reorder)

This is used to sort categories (like religion) based on a numeric value (like average TV hours).

## order levels of one variable by value in another variable
gss_cat %>% 
  group_by(relig) %>% 
  summarise(TV_watchtime_mean = mean(tvhours, na.rm = TRUE)) %>% 
  mutate(relig = fct_reorder(relig, TV_watchtime_mean)) %>% 
  ggplot(aes(TV_watchtime_mean, relig)) +
  geom_point(size = 5, color = "steelblue") +
  theme_light(base_size = 20)

---

4. Modifying Factor Levels (Recoding)

If factor names are unclear or redundant, we can rename them.

## count initial party ID levels
gss_cat %>% 
  count(partyid)

# A tibble: 10 × 2
   partyid                n
   <fct>              <int>
 1 No answer            154
 2 Don't know             1
 3 Other party          393
 4 Strong republican   2314
 5 Not str republican  3032
 6 Ind,near rep        1791
 7 Independent         4119
 8 Ind,near dem        2499
 9 Not str democrat    3690
10 Strong democrat     3490

unique(gss_cat$partyid)  

 [1] Ind,near rep       Not str republican Independent        Not str democrat  
 [5] Strong democrat    Ind,near dem       Strong republican  Other party       
 [9] No answer          Don't know        
10 Levels: No answer Don't know Other party ... Strong democrat

Renaming Levels (fct_recode)

## combine similar levels into "others"
gss_cat %>% 
  mutate(partyid = fct_recode(partyid, 
                              "others" = "Other party",
                              "others" = "No answer",
                              "others" = "Don't know")) %>% 
  count(partyid)

# A tibble: 8 × 2
  partyid                n
  <fct>              <int>
1 others               548
2 Strong republican   2314
3 Not str republican  3032
4 Ind,near rep        1791
5 Independent         4119
6 Ind,near dem        2499
7 Not str democrat    3690
8 Strong democrat     3490

## full recode/rename of levels
gss_cat %>% 
  mutate(partyid = fct_recode(partyid, 
                              "others" = "Other party",
                              "others" = "No answer",
                              "others" = "Don't know",
                              "Not Strong republican" = "Not str republican",
                              "Slightly republican" = "Ind,near rep",
                              "Slightly democrat" = "Ind,near dem",
                              "Not Strong democrat" = "Not str democrat")) %>% 
  count(partyid)

# A tibble: 8 × 2
  partyid                   n
  <fct>                 <int>
1 others                  548
2 Strong republican      2314
3 Not Strong republican  3032
4 Slightly republican    1791
5 Independent            4119
6 Slightly democrat      2499
7 Not Strong democrat    3690
8 Strong democrat        3490

Grouping Levels (fct_collapse)

This is a more efficient way to group multiple levels into single categories simultaneously.

### reorder the similar levels with factor collapse function
gss_cat %>% 
  mutate(partyid = fct_collapse(partyid, 
                                "others" = c("Other party",
                                             "No answer",
                                             "Don't know"),
                                "Republican" = c("Not str republican",
                                                 "Strong republican"),
                                "Democrat" = c("Strong democrat",
                                               "Not str democrat"),
                                "Independent" = c("Ind,near dem", 
                                                  "Ind,near rep"))) %>% 
  count(partyid)

# A tibble: 4 × 2
  partyid         n
  <fct>       <int>
1 others        548
2 Republican   5346
3 Independent  8409
4 Democrat     7180

---

Systematic Checklist (Cheat Sheet):

• Drop Empty Levels: fct_drop()
• Manual Order: fct_relevel()
• Frequency Order: fct_infreq()
• Sort by Numbers: fct_reorder()
• Rename Levels: fct_recode()
• Group/Collapse: fct_collapse()

---

Excellent work! You have mastered the most important tools for handling categorical data in R.

---