Introduction

Module 1.1: Using R

Author

Affiliation

Alex Cardazzi

Old Dominion University

All materials can be found at alexcardazzi.github.io.

Introduction

Students enrolled in this class should have some pre-exposure to both econometrics and R. This course is meant to move students from estimating correlational relationships to causal relationships. Before getting there, we are going to go through an accelerated introduction to R, RStudio, and RMarkdown/Quarto. Since you should already be familiar with the software, this is intended to tighten loose screws rather than build from the ground up. If you find that most of this is new to you, please reach out to me ASAP.

You can find some of the more useful R “cheatsheets” below:

• RStudio Cheatsheet
• Base R Cheatsheet
• dplyr Cheatsheet
• ggplot2 Cheatsheet
• R Syntax Cheatsheet (for translating Base R to tidyverse)
• Regular Expressions
• sf and lubridate Package Cheatsheets

Software

Besides graduating, learning R (or any programming language) is probably one of the most marketable and tangible skills you can pick up in college. Former students have mentioned their earning potentials are clearly higher after learning R (and econometrics). Rather than trying to convince you myself, check out this New York Times article about R from 2009. Note that in the 15 or so years since that article was written, R has grown exponentially. In this class, you will interact with R a lot. However, like learning any language, immersion is the best way, so buckle up.

The first thing you need to do is download and install R onto the machine you’re using. Navigate to this link to do so. After this is completed, you will need to download RStudio. To install RStudio, navigate to this link. When you arrive on the page, you should see two steps. Since you have already downloaded R, you can move onto the second step, which is downloading RStudio.

Why do we need both of these softwares? You should think of R as the brain and RStudio as the body. R, by itself, will work on your machine, but it is much easier to interface with it by using RStudio. After both R and RStudio have been downloaded, open up RStudio. You should see something that has four panels like below:

Screenshot of RStudio

If your RStudio doesn’t look like this RStudio, it’s likely due to differences in themes. Navigate to this link to learn how to change some of these default settings.

The first panel, probably in the lower left corner with a tab labeled “Console”, is where you can view the output of your code. You can write “test” code in the console, but be aware that this code is not saved! In the console, type in print("hello world") and press enter. The output you get should be "hello world". Next, type in 2 + 2 and press enter. Your should get output of 4. If you were to exit out of RStudio and re-open it, this output would be gone.

If you wanted to save this code, you would write it in the top left section with the tab labelled “Untitled1”. This is a .R file. This is where you should write the code you want to keep and/or revisit. It is a good idea to leave yourself notes about the code you’ve written, and you can write yourself “comments” by starting a line with #. As an example, take a look at the following:

# Print the text "hello world"
print("hello world")

The panels to the right will (mainly) be used to display objects in your environment and preview plots you’ve made. We will talk more about this later in the notes.

Lastly, you will be working with Quarto to write homework and other assignments. Quarto is a text editing software like Microsoft Word, except it is built into RStudio. This way, you can produce clean, portable .html documents that include code, output, figures, text, etc. Visit this page for a useful tutorial.

Using R

So, you’ve downloaded the requisite software and you’ve opened RStudio. Now you’d probably like to get to analyzing some data. How do we get from opening RStudio to working with our data?

The first step is going to be checking out RStudio’s current working directory. The working directory is where on your computer RStudio is looking for files. For example, I keep all of my course material in a dropbox folder called: C:/Users/alexc/Dropbox/teaching. In this teaching folder, I have a sub-folder for each semester (e.g. Fall 2023, Spring 2024), and then a sub-folder for each class (e.g. econ400, econ708). For this particular class, my working directory is: C:/Users/acardazz/Dropbox/teaching/Spring 2024/econ400. You can find out what your current working directory is by using getwd(). If your working directory is C:/Users/alexc/Dropbox/teaching, but you want to access a file in C:/Users/alexc/Dropbox/teaching/Spring 2024, then you can use setwd("C:/Users/alexc/Dropbox/teaching/Spring 2024") to change the working directory.

Note: Quarto sets your working directory to where the file itself is saved. So, if your data is in the same folder as your .qmd (Quarto) file, then you do not need to change the working directory.

Next, you will need to load (or read) the data into R. In this class, almost all of the data we’ll work with will be in .csv format. “csv” stands for “comma-separated values”, and is a common, flexible format used across computers. If you have an Excel file, you can simply save it as a .csv, but note that any Excel formulas you have will be removed. If your data is stored in your working directory folder, you can read it by using read.csv("name_of_file.csv"). If your data is not stored in your working directory, you can read it by giving R the full filepath like: read.csv("C:/Users/file/path/to/file/name_of_file.csv"), or if it is a hyperlink: read.csv("https://www.data.com/data/name_of_file.csv"). Practice reading these diamond characteristic data (data; documentation) into R.

If all you did was use read.csv() and the data’s URL, then you likely saw a bunch of stuff get printed to your console. What happened? R did exactly what you told it to – it read the data and that’s it. We cannot access these data because we haven’t told R to hang onto the data yet. To do this, we have to assign the data to a variable name.

Variable names can be anything from df to xyz to the_Most_Important_Data_ever_2. Variable names cannot start with numbers or special characters, and cannot contain spaces (generally). To assign a value to a variable, we use <- or ->, with the arrow pointing to the variable name. If you want to save the data from read.csv(), you’ll need to assign it a variable like this: diamond <- read.csv("write/filepath/here/file.csv").

If you run this line in R, you shouldn’t see any output in the console. What happened? R read the data like before, but then it stored the result so we can access it later. Most of the time, the data you read into R will be in a spreadsheet format with rows and columns. The data structure for these types of data is called data.frame.

WebR

Before moving on to explore data.frames, I want to mention something you’ll see embedded throughout this course. I will be exhibiting code in each module in static code blocks. Many times, these code blocks, sometimes called code chunks, might generate output, plots, both, or nothing. Unfortunately, these code blocks are, for all intensive purposes, set in stone. In other words, besides collapsing/expanding them, you cannot really interact or experiment with them. This probably stifles student curiosity, since you’ll probably want to tweak things as you’re going through the notes.

To address this, I have included WebR chunks into each module’s notes. These chunks will look different from the static chunks, and I encourage you to interact with them. You can write, alter, and execute code inside each chunk, and each WebR chunk will “remember” what you’ve run in other chunks. Go ahead and explore a bit with the chunks below:

While the WebR chunks can “talk” to one another, and the static chunks can talk to one another, there is no communication between the two types of chunks.

# This is a static chunk
# Notice how you cannot modify what's written here.

One more thing before getting back to data.frames: use the following WebR chunk to read in the diamonds data so we can practice with it throughout the notes:

Data Frames

data.frame objects are nothing more than flat spreadsheets. These spreadsheets have dimensions (rows, columns). For example, to get the dimensions of our data.frame we called diamond, we can use dim(diamond). Give this a try in the following code block:

We can also find the number of rows or columns by themselves using nrow() and ncol(). Try these functions out in the code block above, too.

Another way to think of data.frame objects is as collections of named columns (or vectors) all having the same number of rows. We can ask R for the names of the columns by using the colnames() function. We can also use this function to rename columns.

These functions tell us things about the data.frame, but not about the data that’s inside. If you want to look at the data, you can always put the name of the object into the console and press enter. The result will be…the entire dataset. In this case, there are 351 rows, which is likely more than we really want to see. If you do want to see entire data.frame, however, you can either click on the object in your environment tab or you can use the View() function. If you just want to get a quick snapshot, though, you can use head(). This will print just the first six rows. We can also get the last six rows by using tail(). We can control the number of rows that gets printed by adding a seconding argument to the function head() (or tail()). For example, to get the first 10 rows, we can use head(diamond, 10). Taking quick looks at our data is helpful for checking on what we’re working with, so it’s helpful to be familiar with head() and View().

We can also specify which rows in particular we want by giving R a vector of indices. To get the first five rows, we can use the following code: diamond[c(1, 2, 3, 4, 5) , ]. Let’s break this down a bit:

• diamond tells R the data we want to subset
• We use square brackets ([]) to subset things in R. Since data.frame objects have two dimensions (rows, columns), we use a comma between the square brackets to indicate this split.
• c(1, 2, 3, 4, 5) is a vector of row numbers that we want to see. R knows we are talking about rows because we put the vector to the left of the comma between the square brackets. If we put it to the right of the comma, we would have gotten the first five columns instead of rows. Notice how when we leave blank space between the comma and the square bracket, we get all the columns (or rows, if the vector was to the right of the comma).

Practice subsetting in the chunk below:

What if we wanted to access the column containing information about the size (referred to as “carats”) of the diamonds? If we use colnames(diamond), we can see which column corresponds to carats. Once we know the position of the column in the data.frame, we can use the square brackets to subset the data so we’re only left with that specific column. See if you can pick out the column below:

Since the columns have names, however, we can also access the columns directly by using their name following a dollar sign. For example, to do the same thing as we did above, we can also write: diamond$Carat which will result in the same output.

Let’s return to subsetting rows for a second. What if we don’t know the indices of the rows we want to keep? We can use logic to help us subset. For example, suppose we only want to consider diamonds that are smaller than one carat in our analysis. How could we do this? We can ask R, “hey, which diamonds have a value of carat less than 1?” Converting this to code, we can write: diamond$Carat < 1. The output of this code will be a bunch of TRUE and FALSE values. If we put these values in place of the index vector we had before, R will keep only the rows with a TRUE value. Check this out below:

Remember: if you want to make this subset permanent, you’ll have to assign it to a variable/object! If you subset the data without assigning it, R will subset, print, and then go back to the original data. You can assign it to itself, which is like saving over it, if you’re confident that you never want to see the bigger-than-one-carat rows again. You can also save the subset data as something else altogether, which would allow you to keep both the original and subsetted data.

Before we save over the data, let’s think about how to apply multiple conditions to our subset. For example, maybe we want to keep diamonds that are between one and two carats. We can always remove all diamonds that are bigger than two carats, and then remove diamonds smaller than one carats. We could also do this in one step by asking R for the diamonds smaller than one carats and bigger than one carat. See the following code:

# Version 1
diamond <- diamond[diamond$Carat < 2,]
diamond <- diamond[diamond$Carat > 1,]

# Version 2
diamond <- diamond[(diamond$Carat < 2) & (diamond$Carat > 1),]

What if we wanted to subset or filter for certain diamond colors? For example, we want diamonds that have the color "D", "E", or "F". Note that this is different from the previous multiple-condition subset. Before, we said we only wanted rows where condition A and condition B were true. Here, we are saying we want condition A or condition B or condition C. To use and, we use &. To use or, we use |. So, our subset logic would look like the following:

diamond$Color == "D" | diamond$Color == "E" | diamond$Color == "F"

This can be fairly tedious code to write if we have, like, 10 different options we want to include. A better way to write code that follows this pattern is as follows:

diamond$Color %in% c("D", "E", "F")

Of course, you can start getting wild with this, and subset data where the size is within some range of carats and some specific colors.

Data Manipulation

Now that we know some of the basics on how to subset data, let’s talk about how to manipulate data. First, simple arithmetic (adding, subtracting, multiplying, dividing) is pretty easy. R is a vectorized language, which means that if we multiply a vector by 5, each element gets multiplied by 5. This might not seem like a big deal, but this is not the case in every programming language.

To multiply the values in a column by 5, we can use the following code: diamond$Carat * 5. Note that by doing this, you are multiplying the column by 5 but that’s it. To save this change, you need to save over it like this: diamond$Carat <- diamond$Carat * 5.

Instead of saving over the original column, we can create a new column. For example, we could use diamond$Carat_5 <- diamond$Carat * 5. We can also create variables that are functions of multiple columns. For example, we can calculate price per carat by dividing diamond$Carat by diamond$TotalPrice, and saving it as a new variable.

Next, let’s discuss creating data based on logical values. For example, maybe we want to label diamonds that are bigger than 1 carat as “big”. To do this, we can start with the logic: diamond$Carat > 1. Then, when this is true, we want a value of 1 and if false, a value of 0. To do this, we can use ifelse(). In sum, we can use the following code: ifelse(diamond$Carat > 1, 1, 0). Again, if we want to save this, we need to create a new column for it.

Binary variables are sometimes called dummy variables.

Use the following code block to explore creating new columns and/or modifying old columns:

To “reset” the data to its original form, you will need to read it back in from its URL.

Text Data

Often times, data comes in the form of text rather than numeric. If we want to use text as data, we need to learn how to interact with text to turn it into numeric data. The first function you should know is grepl() – this is similar to ctrl + f on Windows. You will pass a character argument followed by a vector argument to grepl(), and it will return TRUE or FALSE depending on if R found the character in each element of the vector.

Diamond clarity has a bit of a strange scale that ranges from “flawless” to “included”. Some of the middle tier clarity categories have the word “very” in them. For example: “very slightly included”. Let’s suppose, for some reason, we want to create a dummy variable equal to one if the diamond’s clarity category contains “very” in it. By looking at the data, you might have noticed that the diamond clarity column was filled with abbreviations. Therefore, we just need to search for the letter "V" rather than the word "very". The code we’ll need is as follows: grepl("V", diamond$Clarity).

While TRUE/FALSE values might help us create dummy variables or subset data, we can also do find-and-replace in R. To do this, we can use gsub(). gsub() works like grepl() in that you provide a “find” character and a “where to look” vector. However, you need to also provide another character that tells R what to replace the find character with. For example, suppose we want to replace "V" with "very". We would do that with the following: gsub("V", "very", diamond$Clarity).

We can also “combine” two variables into one by using paste(). For example, suppose we had a vector of schools skool <- c("ODU", "WVU", "RCNJ") and a vector of states statez <- c("VA", "WV", "NJ"). We could combine these two vectors by using: paste(skool, statez). By default, paste() included a space between the two elements, but you can control this by specifying sep = "..." in paste(), where ... is whatever you want to be included between the elements (use "" for nothing).

As a final note on data manipulation, check out lubridate to learn how to convert date-text to numeric.

Try experimenting with some of these text functions below:

Important Functions

So far, we have seen some of how R works. We manipulate objects (i.e. vectors, data.frames, etc.) with functions (i.e. paste(), head(), ifelse()). R has many built-in functions for us to use to move around and organize data, as well as compute all sorts of statistics. Let’s go over a few of the most important ones.

table()

Now that we’ve created some new data, it might be helpful to summarize it. diamond has 351 observations and it may be difficult to contextualize all of it. For example, we might want to know, “what’s the average price of diamonds?” We can use mean() to calculate the average. We can also calculate the median (median()) and standard deviation (sd()). While these give helpful single number summaries, it might be helpful to check out the entire distribution, too. One easy way to do this is via table(). This function will show you all of the unique values in the data and count how many times each appears. Below, try tabulating the color of diamonds. Then, tabulate the total price of the diamonds.

You will probably get varying results for color and price. Since color is categorical (with few options), the table is clean and easy to read. You can quickly pick out the most frequent color, the second most frequent, etc. Since the prices of diamonds vary so much, it’s unlikely you’ll find two diamonds with exactly the same price. For continuous data like this, it’s helpful to round the data before tabulating it. To round the data, we can use the round() function. Try tabulating the rounded price rather than the raw price.

This probably didn’t help either, huh? Maybe a better idea would be to first convert the price into thousands of dollars by dividing by 1,000, rounding, then multiplying by 1,000. You can do this via: round(diamond$TotalPrice/1000)*1000. Try this out in the code block above. Tabulating this creates a much more intuitive and helpful distribution. We can once again easy pick out where the most diamonds fall price-wise, etc.

We can also visualize tabulations by simply using the plot() function. We will talk more about how to plot in R at the end of these notes, but for instructional purposes, try plotting the distribution of the rounded price data above. After, try getting rid of the rounding, multiplication, and division (or change 1000 to 1) to see what the distribution of the raw data looks like compared to the rounded data.

aggregate()

Another helpful function for summarizing/collapsing data is aggregate(). This function allows us to find, for example, the average size and price by diamond color. We can easily calculate the average price for all diamonds by using mean(), but what if we wanted to find the average price for each diamond color? We would have to use aggregate(). See below for an example:

          #x: these are the variables we want to aggregate
aggregate(list(Carat = diamond$Carat,
               Price = diamond$TotalPrice),
          # by: these are the groups by which to aggregate
          list(Color = diamond$Color),
          # FUN: this is the function we are going to apply
          mean) -> agg
agg

Output

  Color     Carat    Price
1     D 0.8225000 5568.798
2     E 0.7747561 4359.462
3     F 1.0568966 9160.397
4     G 1.1684884 9558.899
5     H 1.2144828 8668.721
6     I 1.2216667 7639.925
7     J 0.7366667 1936.300

There are some nuances about the above code. Notice how the first two arguments are lists. In addition, within these lists, I set names. These names get passed as column names in agg. By using lists, we are able to aggregate multiple columns by other multiple columns. It’d be worthwhile to explore this in one of the WebR chunks.

match()

Another important function you will come across is match(). This is one way to merge two different sets of data. For example, if there is one data.frame with unemployment rate as a variable and another that has GDP as a variable, we might want to combine the two into a single data.frame. Check out the code below to explore how the match function works.

Loops

Next, we will discuss loops in R. One of the most used loop in R is the for loop. This allows us to repeat some chunk of code a set number of times. I know we used aggregate() to find the average price per color type, but for instructional purposes, let’s use loops to replicate the results. As a disclaimer, you should use aggregate() over loops almost always. This is purely for instruction.

To start, we need to tell R how many times we want the loop to execute. To do this, we have to give R a vector to loop over. Sometimes, this can be a vector like 1:10, or it can be c("D", "E", "F"). Then, we need to write the structure of the loop which is usually as follows: for(index in looped_vector). looped_vector can be replaced with c("D", "E", "F"), and index can be replaced with, for example, color. The index is a placeholder variable that will take on different values of the vector being looped over. If we are looping over c("D", "E", "F"), then color would equal "D" the first time through the loop, then "E", and then "F". The code inside the loop will be repeated a total of three times, since there are three elements in the vector being looped over. Let’s now look at what the code would look like.

To start, try simply printing color each time through the loop. Specifically, replace the comment in the loop with print(color) and see what the output looks like. If done correctly, you should get three lines of output. What happened? Even though you printed color three times, the value of color changed each time through the loop. Next, modify the code above to print the average price of diamonds that are the same color as color. See below for a solution.

for(color in c("D", "E", "F")){
  
  print(mean(diamond$TotalPrice[diamond$Color == color]))
}

Output

[1] 5568.798
[1] 4359.462
[1] 9160.397

To do this in a more complete way, we could use all of the possible colors. Of course, we don’t want to have to write out all of the colors. Rather, we can just use unique(diamond$Color).

Custom Functions

Sometimes, we want to create specific functions to use. R has already built in a function to calculate averages (mean()), but maybe we want to write our function to calculate the mean.

To write your own function, you must start with giving the function a name. Since our function is going to calculate averages, let’s call it average. Next, we need to tell R what “arguments” our function will accept. To calculate an average, we really just need a vector of numbers. Let’s use c(1, 8, 8, 10, 4, 9), which we can name anything we’d like. In our function, we will name the vector argument vec.

Mechanically, to calculate an average, we need to add up all the numbers and divide by how many there are. We can use sum() and length() for this. So, inside our function, we want to take the sum of vec and then divide by the length of vec. Note that vec can be any vector of numbers, not just c(1, 8, 8, 10, 4, 9). Take a look below at how I write the function:

To use the function, use average(vec = v) in the chunk above after the function. For practice, try writing your own function for standard deviation.

Packages

R is a free, open source programming language. For our purposes, this just means that other people have written their own functions that we can download and use. Why do we want to use code that others have written? So we don’t re-invent the wheel! Some of the things we’ll learn in this course involve relatively complex algorithms, so we’ll make use of code others have written to make our lives easier.

Collections of “other peoples’ code” are called “packages”. There are two steps to using packages. First, we need to install the package. You should think of this as downloading the code from the internet and onto your personal computer. Second, we need to load the package into our specific R session. We only ever need to install the package once, but we need to load the package every time we open up R and want to use it.

Note: you should not load every package you’ve ever installed every time you open R. This will be tempting, as I often see students with like twelves packages loaded at the beginning of their code, but this is a bad habit. Try your best to get familiar with which packages correspond to which functions, etc. If your code breaks because you haven’t loaded a certain package, it’s fine! Everything is fine. Error messages are not bad and they are not trying to be mean. Learn to read them and understand them. They are doing their best to help you. If you feel stuck and frustrated, congrats, you’re just like everyone else! You should try Googling the error message, emailing me (with your code & error message), etc. Learning to program is hard, which is why you’re doing it in school before you get to a job, etc.

Sorry for the tangent – how do we install and load these packages? To install packages, we will use install.packages(). For example, in a bit we are going to use the scales package, so let’s try installing this on your machine. If you have already done this, you do not have to do it again! Next, to load the scales package into your current R session, use library("scales"). Once you do these two steps, you should be able to access a function called alpha() that was not available to you before loading scales. If you don’t believe me, close out of R, re-open it, and type ?alpha into the console and hit enter. R will likely return No documentation for ‘alpha’ in specified packages and libraries. If you use library("scales") and then type in ?alpha, you should get something pop up in the bottom right panel of RStudio.

I will pre-install packages for you in WebR chunks. Installation is a bit different, so do not use install.packages() in WebR chunks. However, you do need to load libraries in WebR chunks before using them.

You can try this in the below WebR block. Run ?alpha and see the output. Then un-comment the line with library() and run the code again. You should get different output. Then, if you re-comment library(), the result will not change from the run before. This is because you only need to load the library once per session.

Plotting

Another way to summarize data is via visualization. I know it’s cliche, so I won’t say it, but this is an incredibly valuable skill in data science and econometrics. Entire courses are taught on data visualization, so we will only be able to get through the basics in this course. We are going to produce simple but informative figures throughout the semester and hopefully you will get better and more creative over time.

The way R’s graphics work are like layers. You start with a set of axes and add layers to the top of the canvas. You can never add something underneath – only on top. Keep this in mind as we go through the following examples.

In this section, we will start by reading in data on the list prices of Ford Escorts of different years and mileages (link).

Before getting to plotting, it’s a good idea to explore some of the data (figuring out how many rows/columns, what the column names are, what the data looks like, etc.). Use the code block below to poke around a bit. It’s a relatively small dataset, so this shouldn’t be too difficult.

If you check out the data, you might notice that some of the column names are a bit funky. Let’s rename them to "year", "mileage", and "price". Let’s also create a new column that is the price per mile.

Let’s begin by visualizing the relationship between mileage and price. To do this, we can simply use plot(ford$mileage, ford$Price). Note that whichever column you put first will be on the x-axis, and the column you put second will be on the y-axis. Go ahead and use the code above in the chunk below to generate your first plot.

This plot is functional, but it could be much better. Below are some examples of additional arguments accepted by the plot() function.

1. las = 1 rotates the text on the y-axis. Different numbers will rotate it more or less.
2. col sets the colors used in the plot. This can take a vector of colors for multi-colored plots.
3. pch sets the type of point used.
4. cex sets the size of the points. The default is 1.
5. main sets the title of the plot. The default is blank.
6. xlab sets the name of the x-axis. The default is the vector name you gave the function.
7. ylab sets the name of the y-axis. The default is the vector name you gave the function.

Try exploring some of these options by yourself:

Below, I have provided a working example.

plot(ford$mileage, ford$cost_per_mile, las = 1,
     pch = 23, cex = 1.2,
     col = "tomato", main = "Cost vs Mileage",
     xlab = "Mileage", ylab = "Price per Mile")

Plot

Adding Lines

We can also add reference lines with the abline() function. If you want a horizontal line at some value y, you could use abline(h = y). If you wanted the line to be vertical, you would use v = instead of h =. To use abline(), you much first have the plot axes set up. Remember, R plots are just layers.

Colors

We can also make the colors a bit more complex. Remember how we used logic with the diamond data to subset it? We can do the same thing with coloring in graphs. Let’s color the points of older vehicles blue and newer ones red.

# If the Year is less than the mean year, color it "dodgerblue"
plot(ford$mileage, ford$cost_per_mile, las = 1,
     pch = 19, cex = 1.2, main = "Cost vs Mileage",
     col = ifelse(ford$year < mean(ford$year), "dodgerblue", "tomato"),
     xlab = "Mileage", ylab = "Price per Mile")

Plot

Before moving on, it can be helpful to add opacity to what you plot. Sometimes, figures can get busy and crowded, so transparency can help us see where clusters of points are located. To do this, we need to load in the scales package like we did before. Then, when using the col argument, we will use alpha("tomato", 0.3). The first part of this, where we have "tomato", can be a vector of colors. We can even use our ifelse() colors in this slot. The second argument is the opacity, which is a measure of transparency. A value of 1 means completely solid, and a value of 0 means invisible. You can play with opacity in the chunk below:

Legends

Of course, whenever you choose to add some differences in shapes, colors, etc., it’s helpful to add a legend to your plot. To do this, we can use the legend() function. This function accepts a few important arguments:

• x, y: You can specify the exact coordinates of your legend, or you can specify things like: "topleft", "topright", "bottomleft", or "bottomright".
• horiz: this accepts a boolean value, and turns the legend from vertical to horizontal.
• bty: setting this to "n" removes the box around the legend. I always use this option.
• legend: this is the actual text to be displayed in the legend. It accepts a character vector, so if you colored your plot by men and women, you would use c("Men", "Women").
• col: this should match the legend option, but is the corresponding color for each legend element.
• Then, you will need to specify either pch or lty options to tell R if you want to display points or lines next to your legend. The vectors of pch or lty, like col, should match the elements in legend.

Try creating your own legend in the code chunk below:

Points

When generating figures, you will sometimes need to add data from a different source to the same set of axes. As an example, let’s simply plot the data above, but in two steps instead of one.

To do this, we will use points(). This function accepts nearly every argument plot() does, except you are unable to impact the axes/labels of the plot.

Once you get the hang of using plot() and points() in tandem, you’ll find it convenient that points() does not impact the axes. However, to start, this will be annoying. For example, let’s switch the order of the data in plot() and points().

The plot is different because when plot() is setting the axes, it doesn’t know that you’re planning on using points() next. So, it scales the axes so the data fed into plot() “fits” the space. Once again: layers.

To overcome this, we can use the following trick. We are going to set up our axes before we plot anything. We can do this using range(). Try playing with some of the code below:

Of course, this is a lot more coding than the initial plot’s code. The idea of showing you this is that, now, you can always make sure your data “fits”. This is one of the little things that I use constantly, but it took me a long time to figure out.

Lines

Finally, adding lines to a plot is very similar in that one needs to use lines(). To illustrate, we will examine panel data on cigarette consumption by state (documentation).

Read in the data, inspect it a bit, and then plot sales by year.

This figure might be difficult to understand. Let’s trim it down to just a few states.¹ In addition, color the points differently by state.

This plot can still be improved. It’d be a lot more natural to see the data as lines instead of points. To do this, we can use type = "l". Recreate the plot above, but change the plot type.

Notice two things about this plot. First, there’s only a single color. In R, you should think of a line as a single point. R cannot color different parts of line differently, so it will just take the first color (here, it’s 1, which is black). Second, there are these insane diagonal lines across the plot. This is because R wants to connect everything into a single line.

To fix this, we need to use lines() like we used points() before. This is an example case of a time where we’ll want to set up the axes before we plot anything.

# before, I plotted 0, 0
# now, I am simply keeping the data
#   in plot().
# this way, I don't need to set the axes
#   via ylim() and xlim()
plot(cig$year[choose_rows], cig$sales[choose_rows],
     las = 1, type = "n",
     ylab = "Sales", xlab = "Year")
lines(cig$year[cig$state == 1],
      cig$sales[cig$state == 1],
      col = 1)
# Skipping 2 because it doesn't exist
lines(cig$year[cig$state == 3],
      cig$sales[cig$state == 3],
      col = 3)
lines(cig$year[cig$state == 4],
      cig$sales[cig$state == 4],
      col = 4)
lines(cig$year[cig$state == 5],
      cig$sales[cig$state == 5],
      col = 5)
legend("bottomleft", ncol = 2,
       legend = c("State 1", "State 3", "State 4", "State 5"),
       bty = "n", col = c(1, 3, 4, 5), lty = 1)

Plot

This can also be done via looping. Try finishing the following code:

Footnotes

1. Unfortunately, the states seem to just be numbered instead of labeled, so we’ll just pick 1 through 5. In addition, note that there does not seem to be a state number 2.