Module 13 Dataframes

Learning goal

  • Practice exploring, summarizing, and filtering dataframes

 

A vector is the most basic data structure in R, and the other structures are built out of vectors. But, as a data scientist, the most common data structure you will be working with – by far – is a dataframe.

A dataframe, essentially, is a spreadsheet: a dataset with rows and columns, in which each column represents is a vector of the same class of data.

Here is what a dataframe looks like:

In this dataframe, each row pertains to a unique iris plant. The columns contain related information about each individual plant.

Here’s another data.frame, built from scratch, which shows that dataframes are just a group of vectors:

In this command, we used the data.frame() function to combine two vectors into a dataframe with two columns named x and y. R then saved this result in a new variable named df. When we call df, R shows us the dataframe.

The great thing about dataframes is that they allow you to relate different data types to each other.

This dataframe has one column of class character and another of class numeric.

Subsetting & exploring dataframes

To explore dataframes, let’s use a dataset on fuel mileage for all cars sold from 1985 to 2014.

To look at this dataframe in full, you call display it in a separate tab within RStudio using the View() function:

A dataframe has rows of data organized into columns. In this dataframe, each row pertains to a single vehicle make/model – i.e., a single observation. Each column pertains to a single type of data. Columns are named in the header of the dataframe.

All the same useful exploration and subsetting functions that applied to vectors now apply to dataframes. In addition to those functions you already know, let’s add some new functions to your inventory of useful functions.

Subsetting

Recall that dataframes are filtered by row and/or column using this format: dataframe[rows,columns]. To get the third element of the second column, for example, you type dataframe[3,2].

Note that the comma is necessary even if you do not want to specify columns. If you try to type this …

…R will assume you are asking for the third column, not the third row.

To filter a dataframe to multiple values, you can specify vectors for the row and column

Columns can also be called according to their names. Use the $ sign to specify a column.

Note that when you use a $, you will not need to use a comma within your brackets. If you try to run this …

…R will throw a fit.

Also recall that you can use logical tests, which return boolean values TRUE or FALSE, to filter dataframes to rows that meet certain conditions. For example, to filter to only the rows for cars with better than 100 mpg, you can use this syntax:

Or you can write all this in a single line, to be more efficient:

Recall that the logical test is returning a bunch of TRUE’s and FALSE’s, one for each row of vehicles. Only the TRUE rows will be returned.

Summarizing

The same summary functions that you have used for vectors work for the columns in dataframes, since each column is also a vector. Check it out:

You can also use the summary() function, which provides summary statistics for each column in your dataframe:

The function unique() returns unique values within a column:

Finally, the order() function helps you sort a dataframe according to the values in one of its columns.

Reverse the order by wrapping rev() around the order() call:

Creating dataframes

As shown above, to create a new dataframe, use the data.frame() function.

Error in paste(vehicles$make, vehicles$model): object 'vehicles' not found
Error in eval(expr, envir, enclos): object 'my_vehicles' not found

Note how the columns were named in the data.frame() call, and that each column is separated by a comma.

You can also stage an empty dataframe, which sounds useless but will become very useful as you start working with for loops and other higher-order R tools.

To coerce an object into a format that R interprets as a dataframe, use as.dataframe():

Modifying dataframes

Combining dataframes

To bind multiple dataframes together by row, use rbind():

Note that to be combined, two dataframes have to have the exact same number of columns and the exact same column names.

The only exception to this is adding a dataframe with content an empty dataframe. That can work, and that will be helpful in the Deep R modules ahead.

You can also bind multiple dataframes together by column, using cbind():

Note that to be combined, two dataframes have to have the exact same number of rows and the exact same column names.

Adding columns

To create a new column for a dataframe, use the $ symbol and provide the name of the new column:

Altering values

To alter certain values in the dataframe, you can assign new values to a subset of your dataframe.

Here are four ways to do the same thing: upating Isabelle’s X-factor:

Option 1: Subsetting a single column

Option 2: Subsetting both rows and columns

Option 3: Subsetting a column based on a logical test

Option 3: Subsetting row and columns using logical tests

Exercises

Reading for errors

What is wrong with these commands? Why will each of them throw an error if you run them, and how can you fix them?

1. vehicles[1,15,]

2. vecihles[1:5,]

3. vehicles$hwy[15,]

4. vehicles[1:5,1:13]

 

Subsetting and filtering

5. Subset one field according to a logical test: With no more than two lines of code, get the number of Honda cars in the vehicles dataset.

6. Subset one field according to a logical test for a different field: In a single line of code, show the mileages of all the Toyotas in the dataset.

7. Subset a dataframe to a single subgroup: In a single line of code, determine how many differet car makes/models were produced in 1995.

8. Get the mean value for a subgroup of data: What is the average city mileage for Subaru cars in the dataset?

9. Subset a dataframe to only data from between two values: According to this dataset, how many different car makes/models have been produced with highway mileages between 30 and 40 mpg?

10. Subset by removing NAs: Create a new version of the vehicles dataframe that does not have any NAs in the trans column.

 

Creating dataframes

11. Create a vector called people of 5 peoples names from the class.

12. Show with code how many people are in your vector

13. Create another vector called height which is the number of centimeters tall each of those 5 people are.

14. Combine these two vectors into a data frame.

Now let’s create a new object named animals. This is going to be a dataframe with 4 different columns: species, weight (in kg), color, veg (whether or not the animal is a vegetarian / herbivore).

15. Come up with five species to add to your dataframe and list them in a vector named species.

16. Make the other vectors with details about those species in the correct order.

17. Combine these vectors into a dataframe named animals.

 

Altering dataframes

18. Add a column to your animals dataframe named rank, which ranks each animal from your least favorite (0) to your most favorite (5).

19. Now write code to manually switch the ranking for your top two favorite animals.

20. What is the mean weight of the herbivorous animals that you listed, if any?

21. What is the mean weight of the omnivorous/carnivorous animals that you listed?