Data Frames in R

Content

Arrays generalize the dimensional aspect of a matrix and assume only one data mode. Data frames in R generalize the mode of a matrix and allow mode mixing. Data frames with mode mixing are are the most widely used data objects in R.

Creating Data Frames in R

You can create data frames in R several ways:

importData() and read.table() both read data from an external file as a data.frame
data.frame() binds together R objects of various kinds.
as.data.frame() coerces objects of a particular type to objects of class data.frame.

The data.frame() function will create a data frame from existing objects if all columns have a name and equal length:

# Generate raw data
> my.logical <- sample(c(TRUE, FALSE), size = 20, replace = TRUE)
> my.complex <- rnorm(20) + runif(20) * 1i
> my.numeric <- rnorm(20)
> my.matrix <- matrix(rnorm(40), ncol = 2)

# Create a data frame
> my.df2 <- data.frame(my.logical, my.complex, my.numeric, my.matrix)
> my.df2
my.logical my.complex     my.numeric   X      X2
1 TRUE     -0.360+0.287i  -0.820    0.180  -0.201
2 TRUE      0.028+0.149i   1.932    0.049  -1.345
3 FALSE     1.695+0.969i   1.547   -1.431  -0.567
…

# Generate raw data

> my.logical <- sample(c(TRUE, FALSE), size = 20, replace = TRUE)

> my.complex <- rnorm(20) + runif(20) * 1i

> my.numeric <- rnorm(20)

> my.matrix <- matrix(rnorm(40), ncol = 2)

# Create a data frame

> my.df2 <- data.frame(my.logical, my.complex, my.numeric, my.matrix)

> my.df2

my.logical my.complex my.numeric X X2

1 TRUE -0.360+0.287i -0.820 0.180 -0.201

2 TRUE 0.028+0.149i 1.932 0.049 -1.345

3 FALSE 1.695+0.969i 1.547 -1.431 -0.567

…

The names of the input objects are used for the names in the data frame, but the matrix input reverts to the matrix names since multiple columns are supplied. Row names for the data frame are obtained from the first object with a names(), dimnames(), or row.names() functions.

The attributes of the data objects are not lost when they are combined in a data frame. However, character and logical vectors are converted to factors to facilitate data anlysis. To prevent coercion, pass the vector to data.frame() in a call to the I() function, which returns the vector unchanged but with the added class “AsIs”.

> my.data <- data.frame(MPG, Dist, Climb, Day = I(day))

1	> my.data <- data.frame(MPG, Dist, Climb, Day = I(day))

It is also possible to supply matrices and lists when creating data frames. If a matrix is submitted to data.frame(), it is the same as if the columns were supplied as individual objects. If a list is supplied, it is treated as if its components had been supplied individually. In both cases, suitable names are concocted if none are supplied.

Expanding a Data Grid in R

A unique way to create a data frame in R is to create a data grid. Data grids contain all combinations of the input data.

> my.grid <− expand.grid(col1 = 1:3, col2 = LETTERS[1:3])
> my.grid

  col1 col2
1    1    A
2    2    A
3    3    A
4    1    B
5    2    B
6    3    B
7    1    C
8    2    C
9    3    C

> class(my.grid)
“data.frame”

> my.grid <− expand.grid(col1 = 1:3, col2 = LETTERS[1:3])

> my.grid

col1 col2

1 1 A

2 2 A

3 3 A

4 1 B

5 2 B

6 3 B

7 1 C

8 2 C

9 3 C

> class(my.grid)

“data.frame”

Naming Data Frames and Using Names

Column names are defined when data is declared in the data.frame() function. The row.names argument to data.frame() creates row names, assuming the input vector is the same length as the data. Meanwhile, the attach() function can be also used to make the columns of a data frame visible by variable name. The detach() function subsequently cleans up the .Data directory of these additional objects.

Subscripting R Data Frames

Many extraction operators will generate vector output with class numeric. Name extraction maintains object class:

# Extract column X1
> a <− my.df2$X1
> a
[1] -1.73p  1.017 -1.374 -1.846  0.421  1.005  0.067  …
# Interrogate data
> mode(a)
[1] "numeric"
> class(a)
[1] "numeric"

# Extract column X1 using []
> b <− my.df2["X1"]
> b
       X1
1  -1.739
2   1.017
3  -1.374
4  -1.846
5   0.421
6   1.005
7   0.067
 …

# Interrogate data
> mode(b)
[1] "list"
> class(b)
[1] "data.frame"

# Extract column X1

> a <− my.df2$X1

> a

[1] -1.73p 1.017 -1.374 -1.846 0.421 1.005 0.067 …

# Interrogate data

> mode(a)

[1] "numeric"

> class(a)

[1] "numeric"

# Extract column X1 using []

> b <− my.df2["X1"]

> b

1 -1.739

2 1.017

3 -1.374

4 -1.846

5 0.421

6 1.005

7 0.067

…

# Interrogate data

> mode(b)

[1] "list"

> class(b)

[1] "data.frame"

Sorting Data Frames in R

The traditional sort() and rev() functions take a vector and return a vector of sorted values. To sort larger data structures with several variables in parallel (e.g. tied values across columns) use the order() and sort.list() functions:

# Load library data and sort
> library(MASS)
> painters[sort.list(row.names(painters)), ]

       Composition  Drawing  Colour  Expression  School
Albani       14        14      10       6         E
Barocci      14        15       6       10        C
Bassano       6         8      17       0         D
Bellini       4         6      14       0         D
...

# Load library data and sort

> library(MASS)

> painters[sort.list(row.names(painters)), ]

Composition Drawing Colour Expression School

Albani 14 14 10 6 E

Barocci 14 15 6 10 C

Bassano 6 8 17 0 D

Bellini 4 6 14 0 D

...

The function produces a positive integer index vector that will arrange its arguments in increasing order. To put a data frame x in decreasing order, use sort.list(-x).

The function order() generalizes sort.list() to an arbitrary number of arguments. The function also breaks ties across columns. The following example sorts painters by composition count (descending as indicated by the negative) and then by school (ascending):

> painters[order(-painters["Composition"],painters["School"]),]

         Composition Drawing Colour Expression School
Guercino      18        10      10       4       E
Rubens        18        13      17      17       G
Raphael       17        18      12      18       A
Cortona       16        14      12       6       C
Le Brun       16        16       8      16       H
Da Vinci      15        16       4      14       A
Guilio Romano 15        16       4      14       A

> painters[order(-painters["Composition"],painters["School"]),]

Composition Drawing Colour Expression School

Guercino 18 10 10 4 E

Rubens 18 13 17 17 G

Raphael 17 18 12 18 A

Cortona 16 14 12 6 C

Le Brun 16 16 8 16 H

Da Vinci 15 16 4 14 A

Guilio Romano 15 16 4 14 A

All these functions have an argument na.last that determines the handling of missing values. With na.last=NA (the default for sort()), missing values are deleted; with na.last=TRUE (the default for order()), they are put last.

Combining and Modifying Data Frames in R

You can use data.frame() to combine one or more data frames, or use cbind(), rbind() or merge(). In practice, use rbind() only when you have complete data frames. Do not use it in a loop to add one row at a time to a data frame – this is inefficient.

Merging R Data Frames and Redundant Data

The merge() function combines multiple sources with duplicated data, using shared columns. You can specify different combinations using the by, by.x, and by.y arguments:

# Generate Data
> odds.data <− data.frame(label = letters[1:5], odds = seq(1, 9, 2))
> evens.data <− data.frame(label = letters[5:1], evens = seq(2, 10, 2))

> odds.data

   label odds
1     a    1
2     b    3
3     c    5
4     d    7
5     e    9

> evens.data

   label evens
1     e     2
2     d     4
3     c     6
4     b     8
5     a    10

> merge(odds.data, evens.data, by.x = "label")

    label odds evens
1      a    1    10
2      b    3     8
3      c    5     6
4      d    7     4
5      e    9     2

# Generate Data

> odds.data <− data.frame(label = letters[1:5], odds = seq(1, 9, 2))

> evens.data <− data.frame(label = letters[5:1], evens = seq(2, 10, 2))

> odds.data

label odds

1 a 1

2 b 3

3 c 5

4 d 7

5 e 9

> evens.data

label evens

1 e 2

2 d 4

3 c 6

4 b 8

5 a 10

> merge(odds.data, evens.data, by.x = "label")

label odds evens

1 a 1 10

2 b 3 8

3 c 5 6

4 d 7 4

5 e 9 2

The following table summarizes some of the basic rule for combining objects into data frames:

[table id=16 /]

Splitting and Analyzing Data Frames in R

Splitting data frames is a common manipulation. The split() function works by taking the columns to be included in the split and a group definition equal to the columns used to split the data:

# Generate data
> x <− data.frame(odds = seq(1, 9, 2),
                  logical = c(T, F, T, T, F),
                  letters = letters[1:5])
> x

   odds  logical letters
1   1     TRUE       a
2   3     FALSE      b
3   5     TRUE       c
4   7     TRUE       d
5   9     FALSE      e

# Spit data
> split.data <− split(x[, c(1, 3)], x[, 2])
> split.data

$`FALSE`
   odds letters
2    3      b
5    9      e

$`TRUE`
   odds letters
1    1      a
3    5      c
4    7      d

# Generate data

> x <− data.frame(odds = seq(1, 9, 2),

logical = c(T, F, T, T, F),

letters = letters[1:5])

> x

odds logical letters

1 1 TRUE a

2 3 FALSE b

3 5 TRUE c

4 7 TRUE d

5 9 FALSE e

# Spit data

> split.data <− split(x[, c(1, 3)], x[, 2])

> split.data

$`FALSE`

odds letters

2 3 b

5 9 e

$`TRUE`

odds letters

1 1 a

3 5 c

4 7 d

A common use for split() is to create a data structure accepted by boxplot.

Analyzing R Data Frames with by()

It is often more convenient to split a data frame using the by() function. by() takes a data frame and splits it by rows into new data frames subsetted by the values of one or more factors (e.g. INDICES). The indices must be declared as list objects and then passed to function FUN, which is applied to each subset in turn. The resulting data object has class “by” and is manipulated further for pretty printing:

do.call("rbind", as.list(by(iris, list(Species=iris$Species), function(x){
         y <- subset(x, select= -Species)
         apply(y, 2, mean)
       }
)))

       Sepal.Length Sepal.Width Petal.Length Petal.Width
setosa     5.006        3.428       1.462        0.246
versicolor 5.936        2.770       4.260        1.326
virginica  6.588        2.974       5.552        2.026

do.call("rbind", as.list(by(iris, list(Species=iris$Species), function(x){

y <- subset(x, select= -Species)

apply(y, 2, mean)

}

)))

Sepal.Length Sepal.Width Petal.Length Petal.Width

setosa 5.006 3.428 1.462 0.246

versicolor 5.936 2.770 4.260 1.326

virginica 6.588 2.974 5.552 2.026

Analyzing Data Frames in R with aggregate()

The aggregate() function also allows you to partition a data frame or a matrix by one or more grouping vectors, and then apply a function to the resulting columns that returns a single value (e.g. sum() or mean()).

> iris.x <- subset(iris, select= -Species)
> iris.s <- subset(iris, select= Species)
> aggregate(iris.x, iris.s, mean)

       Sepal.Length Sepal.Width Petal.Length Petal.Width
setosa     5.006        3.428       1.462        0.246
versicolor 5.936        2.770       4.260        1.326
virginica  6.588        2.974       5.552        2.026

> iris.x <- subset(iris, select= -Species)

> iris.s <- subset(iris, select= Species)

> aggregate(iris.x, iris.s, mean)

Sepal.Length Sepal.Width Petal.Length Petal.Width

setosa 5.006 3.428 1.462 0.246

versicolor 5.936 2.770 4.260 1.326

virginica 6.588 2.974 5.552 2.026

aggregate() returns a data frame with a factor variable column for each group/level in the index vector, and a column of numeric values from applying the specified function to the subgroup variables in the data frame.

The following list of functions can be used to assess a data frame.

[table id=17 /]

These functions ship with base R. Additional functions from open-source packages will be introduced in the chapter on large data objects.

Back | Next