r-statistics.co by Selva Prabhakaran

R Factors: The Data Type That Trips Up Almost Every R Beginner

A factor is R's way of storing categorical data — variables with a fixed set of possible values like "Male"/"Female", "Low"/"Medium"/"High", or survey responses. Factors look like text but behave like integers, and that dual nature causes most of the confusion.

Factors are one of R's most misunderstood features. They look like character vectors when you print them, but underneath they're integers with labels. This mismatch causes mysterious bugs: as.numeric() returns wrong numbers, sorting doesn't work as expected, and new values get silently turned into NA. This tutorial explains factors clearly and shows you when to use them — and when to avoid them.

What Is a Factor?

A factor stores categorical data — data with a fixed, known set of possible values called levels. Think of a survey question with predefined answer choices:

# Create a factor from a character vector satisfaction <- factor(c("Happy", "Neutral", "Happy", "Unhappy", "Happy", "Neutral")) cat("Factor:", satisfaction, "\n") cat("Levels:", levels(satisfaction), "\n") cat("Class:", class(satisfaction), "\n") # Under the hood: integers! cat("\nUnderlying integers:", unclass(satisfaction), "\n") cat("Level 1 =", levels(satisfaction)[1], "\n") cat("Level 2 =", levels(satisfaction)[2], "\n") cat("Level 3 =", levels(satisfaction)[3], "\n")

  





R stores "Happy" as 1, "Neutral" as 2, "Unhappy" as 3 internally. The text labels are just a mapping from integers to names. This is more memory-efficient for large datasets with repeated categories.



Creating Factors



From a character vector




  

    
# Basic factor — levels are alphabetical by default
colors <- factor(c("red", "blue", "green", "red", "blue"))
cat("Levels:", levels(colors), "\n")  # Alphabetical: blue, green, red

# Specify levels explicitly (controls the order)
sizes <- factor(
  c("Large", "Small", "Medium", "Large", "Small"),
  levels = c("Small", "Medium", "Large")  # Your order
)
cat("Levels:", levels(sizes), "\n")  # Your order: Small, Medium, Large

    

      
      
    

    

  





Ordered factors



For ordinal data (where the order matters), use ordered = TRUE:




  

    
# Ordered factor — levels have a natural order
education <- factor(
  c("PhD", "Bachelor", "Master", "Bachelor", "PhD"),
  levels = c("Bachelor", "Master", "PhD"),
  ordered = TRUE
)

cat("Education:", education, "\n")
cat("Levels:", levels(education), "\n")

# Ordered factors support comparisons
cat("\nPhD > Bachelor:", education[1] > education[2], "\n")
cat("Bachelor < Master:", education[2] < education[3], "\n")

    

      
      
    

    

  





Ordered factors let R know that "PhD" > "Master" > "Bachelor" — useful for ordinal regression and ordered plots.



Controlling which levels exist



You can define levels that don't appear in the data yet:




  

    
# All possible responses, even if not all are present
responses <- factor(
  c("Agree", "Agree", "Neutral"),
  levels = c("Strongly Disagree", "Disagree", "Neutral", "Agree", "Strongly Agree")
)

cat("Data:", responses, "\n")
cat("All levels:", levels(responses), "\n")

# table() shows counts for ALL levels, including zeros
print(table(responses))

    

      
      
    

    

  





This is important for surveys, experiments, and any situation where you need counts for categories that might have zero observations.



Factor Traps (and How to Avoid Them)



Trap 1: as.numeric() on factors



This is the #1 factor trap — it returns level codes, not the values you see:




  

    
# The classic trap
ratings <- factor(c("5", "3", "4", "5", "2"))
cat("Ratings:", ratings, "\n")

# WRONG: gives level codes (1, 2, 3...), not the actual numbers!
wrong <- as.numeric(ratings)
cat("WRONG (level codes):", wrong, "\n")

# RIGHT: convert to character first, then numeric
right <- as.numeric(as.character(ratings))
cat("RIGHT (actual values):", right, "\n")
cat("Mean rating:", mean(right), "\n")

    

      
      
    

    

  





Always convert factor → character → numeric. Never go directly from factor to numeric.



Trap 2: Adding new values



Factors have fixed levels. Adding a value that's not in the levels produces NA:




  

    
colors <- factor(c("red", "blue", "green"))
cat("Levels:", levels(colors), "\n")

# Try to add a new value
colors[4] <- "yellow"
cat("After adding 'yellow':", colors, "\n")  # NA!

# Fix: add the level first
colors <- factor(c("red", "blue", "green"))
levels(colors) <- c(levels(colors), "yellow")
colors[4] <- "yellow"
cat("After adding level first:", colors, "\n")

    

      
      
    

    

  





Trap 3: Sorting and ordering



Factors sort by level order, not alphabetically:




  

    
# Default factor: levels are alphabetical
months <- factor(c("Mar", "Jan", "Feb", "Jan", "Mar"))
cat("Default sort:", sort(months), "\n")  # Alphabetical

# With custom levels: sorts by level order
months <- factor(
  c("Mar", "Jan", "Feb", "Jan", "Mar"),
  levels = c("Jan", "Feb", "Mar")
)
cat("Custom sort:", sort(months), "\n")  # Calendar order!

    

      
      
    

    

  





This is actually a feature — it's how you get months, weekdays, or any custom category to appear in the right order in plots and tables.



Trap 4: Merging/combining factors




  

    
f1 <- factor(c("a", "b"))
f2 <- factor(c("b", "c"))

# c() on factors gives unexpected results!
combined <- c(f1, f2)
cat("c() result:", combined, "\n")       # Numbers, not letters!
cat("Type:", class(combined), "\n")      # Integer!

# Fix: convert to character first
combined_right <- factor(c(as.character(f1), as.character(f2)))
cat("Correct:", combined_right, "\n")
cat("Levels:", levels(combined_right), "\n")

    

      
      
    

    

  





c() strips the factor class and returns the underlying integers. Always convert to character before combining factors.



Modifying Factors



Reorder levels




  

    
# Change the display/sort order without changing the data
sizes <- factor(c("S", "M", "L", "XL", "M", "S"))
cat("Default order:", levels(sizes), "\n")  # Alphabetical: L, M, S, XL

# Reorder
sizes <- factor(sizes, levels = c("S", "M", "L", "XL"))
cat("Custom order:", levels(sizes), "\n")  # S, M, L, XL

# The data is unchanged
cat("Values:", sizes, "\n")
print(table(sizes))

    

      
      
    

    

  





Relabel levels




  

    
# Rename the categories
status <- factor(c("Y", "N", "Y", "Y", "N"))
cat("Before:", levels(status), "\n")

levels(status) <- c("No", "Yes")  # Maps N→No, Y→Yes (alphabetical order!)
cat("After:", status, "\n")
cat("Levels:", levels(status), "\n")

    

      
      
    

    

  





Warning: When relabeling with levels(x) <-, the new names must be in the same order as the current levels (alphabetical by default). Check levels(x) first!



Drop unused levels




  

    
# After filtering, unused levels remain
colors <- factor(c("red", "blue", "green", "red", "blue"))
subset <- colors[colors != "green"]
cat("After filtering:", subset, "\n")
cat("Levels still include green:", levels(subset), "\n")

# Drop unused levels
clean <- droplevels(subset)
cat("After droplevels:", levels(clean), "\n")

    

      
      
    

    

  





The forcats Package (Modern Factor Handling)



The forcats package (part of the tidyverse) provides cleaner functions for common factor operations:




  

    
library(forcats)

# Create a factor
satisfaction <- factor(c("Happy", "Neutral", "Happy", "Unhappy",
                         "Happy", "Neutral", "Unhappy", "Happy"))

# fct_infreq: order by frequency (most common first)
by_freq <- fct_infreq(satisfaction)
cat("By frequency:", levels(by_freq), "\n")
print(table(by_freq))

    

      
      
    

    

  






  

    
library(forcats)

responses <- factor(c("Agree", "Neutral", "Agree", "Disagree",
                      "Strongly Agree", "Agree", "Neutral"))

# fct_relevel: move specific levels to the front
reordered <- fct_relevel(responses, "Strongly Agree", "Agree")
cat("Reordered:", levels(reordered), "\n")

# fct_collapse: combine levels
collapsed <- fct_collapse(responses,
  Positive = c("Strongly Agree", "Agree"),
  Other = c("Neutral", "Disagree")
)
cat("\nCollapsed:", collapsed, "\n")
print(table(collapsed))

    

      
      
    

    

  






  

    
library(forcats)
library(ggplot2)

# fct_rev: reverse level order (useful for horizontal bar charts)
data <- data.frame(
  fruit = factor(c("Apple","Banana","Cherry","Date","Apple","Banana","Apple")),
  count = 1
)

# Default: alphabetical (A at bottom of horizontal bars)
# fct_infreq + fct_rev: most frequent at top
ggplot(data, aes(y = fct_rev(fct_infreq(fruit)))) +
  geom_bar(fill = "steelblue") +
  labs(title = "Fruit Frequency (ordered)", x = "Count", y = NULL) +
  theme_minimal()

    

      
      
    

    

  





forcats cheat sheet




  
    

      Function
      What it does
    

  
  
    

      fct_infreq()
      Order by frequency
    

    

      fct_rev()
      Reverse level order
    

    

      fct_relevel()
      Move levels to front
    

    

      fct_reorder()
      Reorder by another variable
    

    

      fct_collapse()
      Combine multiple levels
    

    

      fct_lump_n()
      Keep top n levels, lump rest into "Other"
    

    

      fct_recode()
      Rename specific levels
    

    

      fct_drop()
      Drop unused levels
    

  




When to Use Factors vs Characters




  
    

      Use factors when...
      Use characters when...
    

  
  
    

      You need a specific display order in plots
      The text has no fixed categories
    

    

      Statistical models need categorical variables
      You're doing string manipulation
    

    

      You want counts for all categories (including zero)
      Categories aren't predefined
    

    

      Data has ordinal categories (Low < Medium < High)
      You're joining/merging data
    

    

      Memory efficiency matters (millions of rows)
      You're new to R (fewer surprises)
    

  




Modern R advice: Use character vectors by default. Convert to factors only when you need specific level ordering (for plots) or when a statistical model requires it. readr::read_csv() and R 4.0+ default to character, not factor.



Practice Exercises



Exercise 1: Survey Analysis




  

    
# Exercise: A survey collected shirt size preferences:
sizes_raw <- c("M", "L", "S", "XL", "M", "S", "L", "M", "XS", "L",
               "M", "S", "L", "XL", "M", "S", "XXL", "M", "L", "S")
# 1. Convert to a factor with levels in size order (XS, S, M, L, XL, XXL)
# 2. Find the most popular size
# 3. Find how many people chose XS or XXL
# 4. Create a bar chart ordered by size (not frequency)

# Write your code below:


    

      
      
    

    

  






Click to reveal solution



  

    
# Solution
library(ggplot2)

sizes_raw <- c("M", "L", "S", "XL", "M", "S", "L", "M", "XS", "L",
               "M", "S", "L", "XL", "M", "S", "XXL", "M", "L", "S")

# 1. Factor with size order
sizes <- factor(sizes_raw, levels = c("XS", "S", "M", "L", "XL", "XXL"))

# 2. Most popular
tab <- table(sizes)
cat("Most popular:", names(which.max(tab)), "with", max(tab), "votes\n")

# 3. XS or XXL count
extreme <- sum(sizes %in% c("XS", "XXL"))
cat("XS or XXL:", extreme, "of", length(sizes), "\n")

# 4. Bar chart in size order
ggplot(data.frame(size = sizes), aes(x = size)) +
  geom_bar(fill = "steelblue") +
  labs(title = "Shirt Size Distribution", x = "Size", y = "Count") +
  theme_minimal()

    

      
      
    

    

  





Explanation: The levels parameter in factor() controls both the sort order and the plot axis order. Without it, sizes would appear alphabetically (L, M, S, XL, XS, XXL), which makes no sense.






Exercise 2: Fix the Factor Bugs




  

    
# Exercise: This code has 3 factor-related bugs. Find and fix them.
# Goal: calculate the average numeric rating from survey data

ratings <- factor(c("5", "3", "4", "5", "2", "4", "3", "5"))

# Bug 1: Wrong numeric conversion
avg <- mean(as.numeric(ratings))
cat("Average rating:", avg, "\n")

# Bug 2: Can't add a new category
# ratings[9] <- "1"

# Bug 3: Combining two factor vectors
batch1 <- factor(c("A", "B"))
batch2 <- factor(c("C", "D"))
# all_batches <- c(batch1, batch2)

# Write your fixes below:


    

      
      
    

    

  






Click to reveal solution



  

    
# Solution
ratings <- factor(c("5", "3", "4", "5", "2", "4", "3", "5"))

# Bug 1 fix: factor → character → numeric
avg <- mean(as.numeric(as.character(ratings)))
cat("Average rating:", avg, "\n")

# Bug 2 fix: add level before adding value
levels(ratings) <- c(levels(ratings), "1")
ratings[9] <- "1"
cat("With new value:", ratings, "\n")

# Bug 3 fix: convert to character before combining
batch1 <- factor(c("A", "B"))
batch2 <- factor(c("C", "D"))
all_batches <- factor(c(as.character(batch1), as.character(batch2)))
cat("Combined:", all_batches, "\n")
cat("Levels:", levels(all_batches), "\n")

    

      
      
    

    

  





Explanation: All three bugs stem from the same root cause: factors are integers wearing text masks. as.numeric() exposes the integers. Adding new values without new levels creates NA. c() on factors strips the factor class.






Exercise 3: Ordered Analysis




  

    
# Exercise: Create an ordered factor for education levels and use it:
# Levels: "High School" < "Bachelor" < "Master" < "PhD"
# Data: c("Bachelor", "PhD", "Master", "Bachelor", "High School",
#         "Master", "PhD", "Bachelor", "Master", "High School")
#
# 1. Create the ordered factor
# 2. Count people at each level
# 3. Find how many have at least a Master's degree
# 4. What percentage have more than a Bachelor's?

# Write your code below:


    

      
      
    

    

  






Click to reveal solution



  

    
# Solution
edu_data <- c("Bachelor", "PhD", "Master", "Bachelor", "High School",
              "Master", "PhD", "Bachelor", "Master", "High School")

edu <- factor(edu_data,
  levels = c("High School", "Bachelor", "Master", "PhD"),
  ordered = TRUE)

# 1. Ordered factor
cat("Levels:", levels(edu), "\n")
cat("Is ordered:", is.ordered(edu), "\n")

# 2. Count at each level
cat("\nDistribution:\n")
print(table(edu))

# 3. At least Master's
at_least_master <- sum(edu >= "Master")
cat("\nAt least Master's:", at_least_master, "\n")

# 4. More than Bachelor's
above_bachelor <- sum(edu > "Bachelor")
pct <- round(above_bachelor / length(edu) * 100, 1)
cat("Above Bachelor's:", above_bachelor, "(", pct, "%)\n")

    

      
      
    

    

  





Explanation: Ordered factors support >, <, >=, <= comparisons. edu >= "Master" returns TRUE for Master and PhD. This is impossible with regular (unordered) factors or character vectors.






Summary




  
    

      Operation
      Code
      Notes
    

  
  
    

      Create
      factor(x)
      Levels alphabetical by default
    

    

      Custom levels
      factor(x, levels = c(...))
      Your order
    

    

      Ordered
      factor(x, levels, ordered = TRUE)
      Enables <, >
    

    

      Get levels
      levels(f)
      Character vector
    

    

      Relabel
      levels(f) <- c("new1", "new2")
      Must match order
    

    

      Reorder
      factor(f, levels = new_order)
      Doesn't change data
    

    

      Drop unused
      droplevels(f)
      After filtering
    

    

      To character
      as.character(f)
      Safe conversion
    

    

      To numeric
      as.numeric(as.character(f))
      Never skip as.character
    

    

      Combine
      factor(c(as.character(f1), as.character(f2)))
      Convert first
    

  




The golden rules:


    
  
  1. Use factors for ordered categories and plot axes
    2. 
  
  2. Use characters as the default for text data
    3. 
  
  3. Never convert factor → numeric directly
    4. 
  
  4. Always convert to character before combining factors
    5. 




FAQ



Why do factors exist at all?



Historical and practical reasons. In early R, factors saved memory (storing integers instead of repeated strings). Today, they're still essential for: (1) controlling plot axis order, (2) ensuring all categories appear in tables and models, (3) ordinal data where the order matters, (4) statistical models that need categorical variables.



Should I use factors or characters with ggplot2?



ggplot2 works with both. Use characters for quick plots. Convert to factor when you need a specific axis order: factor(x, levels = c("Jan", "Feb", "Mar")).



What changed in R 4.0 regarding factors?



Before R 4.0, data.frame() and read.csv() converted strings to factors by default (stringsAsFactors = TRUE). Since R 4.0, the default is FALSE — strings stay as characters. This eliminated the most common source of factor confusion.



How do I reorder bars in a ggplot2 bar chart?



Use forcats::fct_reorder() to order by a numeric variable, or forcats::fct_infreq() to order by frequency. For manual order, set levels: factor(x, levels = c("first", "second", "third")).



What's Next?



Factors complete your understanding of R's type system. Related tutorials:



    
  
  1. R Project Structure — organize your R work properly
    2. 
  
  2. Data Wrangling with dplyr — modern data manipulation
    3. 
  
  3. ggplot2 Visualization — where factors really shine for controlling plot elements
    4. 




        


        
      


      

        

        
© 2016-17 Selva Prabhakaran. Powered by jekyll,
          knitr, and
          pandoc.
          This work is licensed under the Creative Commons License.