typeof returns integer for something that is clearly a factor

If what you wanted to know was "what class was held by a vector?" then use class. If you wanted to test "whether a vector was a factor?" then use is.factor.

The value returned by typeof being integer for factors is a language feature that confused me as well in my early days of R programming. The typeof function is giving information that's at a "lower" level of abstraction. Factor variables (and also Dates) are stored as integers. Datetimes are stored as numeric. Learn to use class or str rather than typeof (or mode). They give more useful information. You can look at the full "structure" of a factor variable with dput:

 dput( factor( rep( letters[1:5], 2) ) )
# structure(c(1L, 2L, 3L, 4L, 5L, 1L, 2L, 3L, 4L, 5L), 
            .Label = c("a", "b", "c", "d", "e"), class = "factor")

The character values that are usually thought of as the factor values are actually stored in an attribute (which is what "levels" returns), while the "main" part of the variable is a set of integer indices pointing to various level "attributes"), named .Label, so mode returns "numeric" and typeof returns "integer". For this reason one usually needs to use as.character that will coerce to what most people think of as "factors", namely their character representations.


More on str - the surprising information for me was it's an abbreviation of "structure" not "string". It can be clearly seen in the bottommost example how str command is capturing it subjectively clearer than dput, naming it “Factor w/ N levels”:

str(head(abalone$Age, 5))
 Factor w/ 3 levels "Mid","Old","Yng": 2 3 1 1 3

Thank you for asking this question, as I've found data types in R confusing and ran into the same issue while processing the Abalone dataset from UCI Machine Learning Repository. I've continued on with the research following the reply by IRTFM. It eventually helped me understand the typing and hopefully could help someone else. I found this resource helpful on understanding R data types: R-supp-data-structures

What I've observed while processing the data.frame from Abalon dataset:

  1. running lapply function on the "Age" column of the data.frame is resulting in a "list" of "character" type objects - due to the lapply property always returning a list even if in this case it could be an atomic vector
  2. further applying unlist function on the "Age" column of the data.frame is resulting in an "atomic vector" of "character" type object
  3. afer encoding vector as a factor we get a "factor" class object

The code example:

#
# Understanding datatypes while processing Abalone dataset  
#
download.file('http://archive.ics.uci.edu/ml/machine-learning-databases/abalone/abalone.data', 'abalone.data')
abalone  = read.table("abalone.data", header = FALSE, sep=",", na.strings= "*")

# name columns of a data.frame object
colnames(abalone) <- c('Sex', 'Length','Diameter','Height','Whole w.', 'Shucked w.', 'Viscera w.','Shell w.','Rings')
dput(head(abalone, 1))

# discretize numeric rings to three ranges of an abalone age
additiveRingsToAgeConst = 1.5;

abalone$Age = lapply(abalone[,'Rings'] + additiveRingsToAgeConst, function (x) {
  if (x > 11.5)     {"Old"}
  else if (x > 9.5) {"Mid"}
  else              {"Yng"}   
})

# 1. running lapply function on the "Age" column of the data.frame is resulting in a "list" of "character" type objects
dput(head(abalone$Age, 5))
str(head(abalone$Age, 5))
# 2. further applying unlist function on the "Age" column of the data.frame is resulting in an "atomic vector" of "character" type object
abalone$Age = unlist(abalone$Age);
dput(head(abalone$Age, 5))
str(head(abalone$Age, 5))
# 3. afer encoding vector as a factor we get a "factor" class object
abalone$Age = as.factor(abalone$Age)
dput(head(abalone$Age, 5))
str(head(abalone$Age, 5))

Code execution results:

> # 1. running lapply function on the "Age" column of 
  #    the data.frame is resulting in a "list" of "character" type objects
> dput(head(abalone$Age, 5))
list("Old", "Yng", "Mid", "Mid", "Yng")
> str(head(abalone$Age, 5))
List of 5
 $ : chr "Old"
 $ : chr "Yng"
 $ : chr "Mid"
 $ : chr "Mid"
 $ : chr "Yng"

> # 2. further applying unlist function on the "Age" column of the data.frame 
  #    is resulting in an "atomic vector" of "character" type object
> abalone$Age = unlist(abalone$Age);
> dput(head(abalone$Age, 5))
c("Old", "Yng", "Mid", "Mid", "Yng")
> str(head(abalone$Age, 5))
 chr [1:5] "Old" "Yng" "Mid" "Mid" "Yng"

> # 3. afer encoding vector as a factor we get a "factor" class object
> abalone$Age = as.factor(abalone$Age)
> dput(head(abalone$Age, 5))
structure(c(2L, 3L, 1L, 1L, 3L), .Label = c("Mid", "Old", "Yng"
), class = "factor")
> str(head(abalone$Age, 5))
 Factor w/ 3 levels "Mid","Old","Yng": 2 3 1 1 3

Tags:

R

Recent Posts