Exploring Data Frames


Teaching: 20 min
Exercises: 10 min
  • How can I manipulate a data frame?

  • Add and remove rows or columns.

  • Remove rows with NA values.

  • Append two data frames.

  • Understand what a factor is.

  • Convert a factor to a character vector and vice versa.

  • Display basic properties of data frames including size and class of the columns, names, and first few rows.

At this point, you’ve seen it all: in the last lesson, we toured all the basic data types and data structures in R. Everything you do will be a manipulation of those tools. But most of the time, the star of the show is the data frame—the table that we created by loading information from a csv file. In this lesson, we’ll learn a few more things about working with data frames.

Adding columns and rows in data frames

We already learned that the columns of a data frame are vectors, so that our data are consistent in type throughout the columns. As such, if we want to add a new column, we can start by making a new vector:

age <- c(2, 3, 5)
    coat weight likes_string
1 calico    2.1            1
2  black    5.0            0
3  tabby    3.2            1

We can then add this as a column via:

cbind(cats, age)
    coat weight likes_string age
1 calico    2.1            1   2
2  black    5.0            0   3
3  tabby    3.2            1   5

Note that if we tried to add a vector of ages with a different number of entries than the number of rows in the dataframe, it would fail:

age <- c(2, 3, 5, 12)
cbind(cats, age)
Error in data.frame(..., check.names = FALSE): arguments imply differing number of rows: 3, 4
age <- c(2, 3)
cbind(cats, age)
Error in data.frame(..., check.names = FALSE): arguments imply differing number of rows: 3, 2

Why didn’t this work? Of course, R wants to see one element in our new column for every row in the table:

[1] 3
[1] 2

So for it to work we need to have nrow(cats) = length(age). Let’s overwite the content of cats with our new data frame.

age <- c(2, 3, 5)
cats <- cbind(cats, age)

Now how about adding rows? We already know that the rows of a data frame are lists:

newRow <- list("tortoiseshell", 3.3, TRUE, 9)
cats <- rbind(cats, newRow)
Warning in `[<-.factor`(`*tmp*`, ri, value = "tortoiseshell"): invalid
factor level, NA generated

Looks like our attempt to use the rbind() function returns a warning. Recall that, unlike errors, warnings do not necessarily stop a function from performing its intended action. You can confirm this by taking a look at the cats data frame.

    coat weight likes_string age
1 calico    2.1            1   2
2  black    5.0            0   3
3  tabby    3.2            1   5
4   <NA>    3.3            1   9

Notice that not only did we successfully add a new row, but there is NA in the column coats where we expected “tortoiseshell” to be. Why did this happen?


For an object containing the data type factor, each different value represents what is called a level. In our case, the factor “coat” has 3 levels: “black”, “calico”, and “tabby”. R will only accept values that match one of the levels. If you add a new value, it will become NA.

The warning is telling us that we unsuccessfully added “tortoiseshell” to our coat factor, but 3.3 (a numeric), TRUE (a logical), and 9 (a numeric) were successfully added to weight, likes_string, and age, respectively, since those variables are not factors. To successfully add a cat with a “tortoiseshell” coat, add “tortoiseshell” as a possible level of the factor:

[1] "black"  "calico" "tabby" 
levels(cats$coat) <- c(levels(cats$coat), "tortoiseshell")
cats <- rbind(cats, list("tortoiseshell", 3.3, TRUE, 9))

Alternatively, we can change a factor into a character vector; we lose the handy categories of the factor, but we can subsequently add any word we want to the column without babysitting the factor levels:

'data.frame':	5 obs. of  4 variables:
 $ coat        : Factor w/ 4 levels "black","calico",..: 2 1 3 NA 4
 $ weight      : num  2.1 5 3.2 3.3 3.3
 $ likes_string: int  1 0 1 1 1
 $ age         : num  2 3 5 9 9
cats$coat <- as.character(cats$coat)
'data.frame':	5 obs. of  4 variables:
 $ coat        : chr  "calico" "black" "tabby" NA ...
 $ weight      : num  2.1 5 3.2 3.3 3.3
 $ likes_string: int  1 0 1 1 1
 $ age         : num  2 3 5 9 9

Challenge 1

Let’s imagine that 1 cat year is equivalent to 7 human years.

  1. Create a vector called human_age by multiplying cats$age by 7.
  2. Convert human_age to a factor.
  3. Convert human_age back to a numeric vector using the as.numeric() function. Now divide it by 7 to get the original ages back. Explain what happened.

Solution to Challenge 1

  1. human_age <- cats$age * 7
  2. human_age <- factor(human_age). as.factor(human_age) works just as well.
  3. as.numeric(human_age) yields 1 2 3 4 4 because factors are stored as integers (here, 1:4), each of which is associated with a label (here, 28, 35, 56, and 63). Converting the factor to a numeric vector gives us the underlying integers, not the labels. If we want the original numbers, we need to convert human_age to a character vector and then to a numeric vector (why does this work?). This comes up in real life when we accidentally include a character somewhere in a column of a .csv file supposed to only contain numbers, and forget to set stringsAsFactors=FALSE when we read in the data.

Removing rows

We now know how to add rows and columns to our data frame in R—but in our first attempt to add a “tortoiseshell” cat to the data frame we have accidentally added a garbage row:

           coat weight likes_string age
1        calico    2.1            1   2
2         black    5.0            0   3
3         tabby    3.2            1   5
4          <NA>    3.3            1   9
5 tortoiseshell    3.3            1   9

We can ask for a data frame minus this offending row:

cats[-4, ]
           coat weight likes_string age
1        calico    2.1            1   2
2         black    5.0            0   3
3         tabby    3.2            1   5
5 tortoiseshell    3.3            1   9

Notice the comma with nothing after it to indicate that we want to drop the entire fourth row.

Note: we could also remove both new rows at once by putting the row numbers inside of a vector: cats[c(-4,-5), ]

Alternatively, we can drop all rows with NA values:

           coat weight likes_string age
1        calico    2.1            1   2
2         black    5.0            0   3
3         tabby    3.2            1   5
5 tortoiseshell    3.3            1   9

Let’s reassign the output to cats, so that our changes will be permanent:

cats <- na.omit(cats)

Removing columns

We can also remove columns in our data frame. What if we want to remove the column “age”. We can remove it in two ways, by variable number or by index.

           coat weight likes_string
1        calico    2.1            1
2         black    5.0            0
3         tabby    3.2            1
5 tortoiseshell    3.3            1

Notice the comma with nothing before it, indicating we want to keep all of the rows.

Alternatively, we can drop the column by using the index name.

drop <- names(cats) %in% c("age")
           coat weight likes_string
1        calico    2.1            1
2         black    5.0            0
3         tabby    3.2            1
5 tortoiseshell    3.3            1

Appending to a data frame

The key to remember when adding data to a data frame is that columns are vectors and rows are lists. We can also glue two data frames together with rbind:

cats <- rbind(cats, cats)
            coat weight likes_string age
1         calico    2.1            1   2
2          black    5.0            0   3
3          tabby    3.2            1   5
5  tortoiseshell    3.3            1   9
11        calico    2.1            1   2
21         black    5.0            0   3
31         tabby    3.2            1   5
51 tortoiseshell    3.3            1   9

But now the row names are unnecessarily complicated. We can remove the rownames, and R will automatically re-name them sequentially:

rownames(cats) <- NULL
           coat weight likes_string age
1        calico    2.1            1   2
2         black    5.0            0   3
3         tabby    3.2            1   5
4 tortoiseshell    3.3            1   9
5        calico    2.1            1   2
6         black    5.0            0   3
7         tabby    3.2            1   5
8 tortoiseshell    3.3            1   9

Challenge 2

You can create a new data frame right from within R with the following syntax:

df <- data.frame(id = c("a", "b", "c"),
                 x = 1:3,
                 y = c(TRUE, TRUE, FALSE),
                 stringsAsFactors = FALSE)

Make a data frame that holds the following information for yourself:

  • first name
  • last name
  • lucky number

Then use rbind to add an entry for the people sitting beside you. Finally, use cbind to add a column with each person’s answer to the question, “Is it time for coffee break?”

Solution to Challenge 2

df <- data.frame(first = c("Grace"),
                 last = c("Hopper"),
                 lucky_number = c(0),
                 stringsAsFactors = FALSE)
df <- rbind(df, list("Marie", "Curie", 238) )
df <- cbind(df, coffeetime = c(TRUE,TRUE))

Realistic example

So far, you have seen the basics of manipulating data frames with our cat data; now let’s use those skills to digest a more realistic dataset. Let’s read in the gapminder dataset that we downloaded previously:

gapminder <- read.csv("data/gapminder_data.csv")

Miscellaneous Tips

  • Another type of file you might encounter are tab-separated value files (.tsv). To specify a tab as a separator, use "\\t" or read.delim().

  • Files can also be downloaded directly from the Internet into a local folder of your choice onto your computer using the download.file function. The read.csv function can then be executed to read the downloaded file from the download location, for example,

download.file("https://raw.githubusercontent.com/swcarpentry/r-novice-gapminder/gh-pages/_episodes_rmd/data/gapminder_data.csv", destfile = "data/gapminder_data.csv")
gapminder <- read.csv("data/gapminder_data.csv")
  • Alternatively, you can also read in files directly into R from the Internet by replacing the file paths with a web address in read.csv. One should note that in doing this no local copy of the csv file is first saved onto your computer. For example,
gapminder <- read.csv("https://raw.githubusercontent.com/swcarpentry/r-novice-gapminder/gh-pages/_episodes_rmd/data/gapminder_data.csv")
  • You can read directly from excel spreadsheets without converting them to plain text first by using the readxl package.

Let’s investigate gapminder a bit; the first thing we should always do is check out what the data looks like with str:

'data.frame':	1704 obs. of  6 variables:
 $ country  : Factor w/ 142 levels "Afghanistan",..: 1 1 1 1 1 1 1 1 1 1 ...
 $ year     : int  1952 1957 1962 1967 1972 1977 1982 1987 1992 1997 ...
 $ pop      : num  8425333 9240934 10267083 11537966 13079460 ...
 $ continent: Factor w/ 5 levels "Africa","Americas",..: 3 3 3 3 3 3 3 3 3 3 ...
 $ lifeExp  : num  28.8 30.3 32 34 36.1 ...
 $ gdpPercap: num  779 821 853 836 740 ...

We can also examine individual columns of the data frame with our typeof function:

[1] "integer"
[1] "integer"
 Factor w/ 142 levels "Afghanistan",..: 1 1 1 1 1 1 1 1 1 1 ...

We can also interrogate the data frame for information about its dimensions; remembering that str(gapminder) said there were 1704 observations of 6 variables in gapminder, what do you think the following will produce, and why?

[1] 6

A fair guess would have been to say that the length of a data frame would be the number of rows it has (1704), but this is not the case; remember, a data frame is a list of vectors and factors:

[1] "list"

When length gave us 6, it’s because gapminder is built out of a list of 6 columns. To get the number of rows and columns in our dataset, try:

[1] 1704
[1] 6

Or, both at once:

[1] 1704    6

We’ll also likely want to know what the titles of all the columns are, so we can ask for them later:

[1] "country"   "year"      "pop"       "continent" "lifeExp"   "gdpPercap"

At this stage, it’s important to ask ourselves if the structure R is reporting matches our intuition or expectations; do the basic data types reported for each column make sense? If not, we need to sort any problems out now before they turn into bad surprises down the road, using what we’ve learned about how R interprets data, and the importance of strict consistency in how we record our data.

Once we’re happy that the data types and structures seem reasonable, it’s time to start digging into our data proper. Check out the first few lines:

      country year      pop continent lifeExp gdpPercap
1 Afghanistan 1952  8425333      Asia  28.801  779.4453
2 Afghanistan 1957  9240934      Asia  30.332  820.8530
3 Afghanistan 1962 10267083      Asia  31.997  853.1007
4 Afghanistan 1967 11537966      Asia  34.020  836.1971
5 Afghanistan 1972 13079460      Asia  36.088  739.9811
6 Afghanistan 1977 14880372      Asia  38.438  786.1134

Challenge 3

It’s good practice to also check the last few lines of your data and some in the middle. How would you do this?

Searching for ones specifically in the middle isn’t too hard but we could simply ask for a few lines at random. How would you code this?

Solution to Challenge 3

To check the last few lines it’s relatively simple as R already has a function for this:

tail(gapminder, n = 15)

What about a few arbitrary rows just for sanity (or insanity depending on your view)?

Tip: There are several ways to achieve this.

The solution here presents one form using nested functions. i.e. a function passed as an argument to another function. This might sound like a new concept but you are already using it in fact. Remember my_dataframe[rows, cols] will print to screen your data frame with the number of rows and columns you asked for (although you might have asked for a range or named columns for example). How would you get the last row if you don’t know how many rows your data frame has? R has a function for this. What about getting a (pseudorandom) sample? R also has a function for this.

gapminder[sample(nrow(gapminder), 5), ]

To make sure our analysis is reproducible, we should put the code into a script file so we can come back to it later.

Challenge 4

Go to file -> new file -> R script, and write an R script to load in the gapminder dataset. Put it in the scripts/ directory and add it to version control.

Run the script using the source function, using the file path as its argument (or by pressing the “source” button in RStudio).

Solution to Challenge 4

The contents of scripts/load-gapminder.R:

download.file("https://raw.githubusercontent.com/swcarpentry/r-novice-gapminder/gh-pages/_episodes_rmd/data/gapminder_data.csv", destfile = "data/gapminder_data.csv")
gapminder <- read.csv(file = "data/gapminder_data.csv")

To run the script and load the data into the gapminder variable:

source(file = "scripts/load-gapminder.R")

Challenge 5

Read the output of str(gapminder) again; this time, use what you’ve learned about factors, lists and vectors, as well as the output of functions like colnames and dim to explain what everything that str prints out for gapminder means. If there are any parts you can’t interpret, discuss with your neighbors!

Solution to Challenge 5

The object gapminder is a data frame with columns

  • country and continent are factors.
  • year is an integer vector.
  • pop, lifeExp, and gdpPercap are numeric vectors.

Key Points

  • Use cbind() to add a new column to a data frame.

  • Use rbind() to add a new row to a data frame.

  • Remove rows from a data frame.

  • Use na.omit() to remove rows from a data frame with NA values.

  • Use levels() and as.character() to explore and manipulate factors.

  • Use str(), nrow(), ncol(), dim(), colnames(), rownames(), head(), and typeof() to understand the structure of a data frame.

  • Read in a csv file using read.csv().

  • Understand what length() of a data frame represents.