B Appendix B: Homework

This section provides the homework assignments for the course.

B.1 Homework #1

Due date: Sept. 13 by midnight

For this assignment, you will submit the assignment to me by email by the due date. You should include one file in your submission:

  1. A Word document with seven paragraphs. Each paragraph should be headed with the name of one swirl lesson and the body of the paragraph should describe that lesson and what you learned from it.

For this homework assignment, you’ll be working through a few swirl lessons that are relevant to the material we’ve covered so far. Swirl is a platform that helps you learn R in R—you can complete the lessons right in your R console.

Depending on your familiarity with R, you can either work through seven lessons of your choice in the R Programming: The basics of programming in R and Getting and Cleaning Data courses (suggested lessons are listed further below) (Option #1), or you can work through seven lessons of your choice taken from any number of swirl’s available courses (Option #2) .

For each lesson completed, please write a few sentences that cover: 1. A summary of the topic(s) covered in that lesson, and 2. The most interesting thing that you learned from that lesson. Turn in a hardcopy of this (with your first and last name at the top) during class on the due date.

To begin, you’ll first need to install the swirl package:

install.packages("swirl")

If you’ve never run swirl() before, you will be prompted to install a course. You can do that with the install_course function. For example, to install the R Programming course, you would run:

library(swirl)
install_course("R Programming")

Once you’ve installed a course, every time you enter the swirl environment with swirl(), R Progamming should show up as a course option to select. You can enter R Programming to start lessons in that course by typing the number in fromt of it when you run swirl().

Once you have at least one course installed, you call the swirl() function to enter the interactive platform in RStudio. The console will take you through a few prompts: you’ll give swirl a name to call you, and take a look at some commands that are useful in the swirl environment. Those commands are listed further below.

library(swirl)
swirl()

After calling swirl(), you may be prompted to clear your workspace variables by running rm=(list=ls()). Running this code will clear any variables you already have saved in your global environment. While swirl recommends that you do this, it’s not necessary.

Some of these lessons complement online courses through Coursera, so sometimes you will be asked after you complete a lesson if you want to report your results to Coursera. You should select “No” for that option each time.

B.1.1 Option 1

For Option 1 of this homework, you will need to work through seven of the 15 available lessons in the R Programming course. Here are some suggestions for particularly uesful lessons that you could choose (the lesson number within the course is in parentheses):

R Programming course:

  • Basic Building Blocks (1)
  • Sequences of Numbers (3)
  • Vectors (4)
  • Missing Values (5)
  • Subsetting Vectors (6)
  • Logic (8)
  • Looking at Data (12)
  • Dates and Times (14)

Getting and Cleaning Data course:

  • Manipulating Data with dplyr (1)
  • Grouping and Chaining with dplyr (2)
  • Dates and Times with lubridate (4)

Each lesson should take at most 10–15 minutes, but some are much shorter. You can complete the lessons in any order you want, but you may find it easiest to start with the lowest-numbered lessons and work your way up, in the order we’ve listed the lessons here.

You’ll be able to get started on some of these lessons after your first day in class (“Basic Building Blocks”“, for example), but others cover topics that we’ll get to in weeks 2 and 3. Whether or not we’ve covered a swirl topic in class, you should be able to successfully work through the lesson. At the end of each lesson, you may be asked if you would like to receive credit for completing this course on Coursera.org. Always choose”no" for this option.

Again, you’ll need to compose and turn in a few sentences for each lesson. Make sure to include a summary of what each lesson was about, and the most interesting thing about that lesson.

B.1.2 Option 2

If you’re already somewhat familiar with R, you might want to choose your seven lessons from other swirl courses instead of or in addition to those available in the R Programming and Getting and Cleaning Data courses.

Check out the list of available Swirl Courses to see which ones you would like to install and check out available lessons for. For example, to choose a lesson in the Exploratory Data Analysis course, you would run:

library(swirl)
install_course("Exploratory Data Analysis")
swirl()

After entering the Exploratory Data Analysis course, you could choose from any one of its available lessons.

In your written summary for each lesson (again, a few sentences that cover a summary of the lesson and the most interesting thing you learned), make sure to specify which course each lesson you completed was from.

B.1.3 Special swirl commands

In the swirl environment, knowing about the following commands will be helpful:

  • Within each lesson, the prompt ... indicates that you should hit Enter to move on to the next section.
  • skip(): skip the current question.
  • play(): temporarily exit swirl. It can be useful during a swirl lesson to play around in the R console to try things out.
  • nxt(): regain swirl’s attention after play()ing around in the console.
  • main(): return to swirl’s main menu.
  • bye(): exit swirl. Swirl will save your progress if you exit in the middle of a lesson. You can also hit the Esc. key to exit. (To re-enter swirl, run swirl(). In a new R session you will have to first load the swirl library: library(swirl).)

B.1.3.1 For fun

While they aren’t required for class, you should consider trying out some other swirl lessons later in the course. You can look through the course directory to see what other courses and lessons are available. For the first part of our course, you might find the “Exploratory Data Analysis” course helpful. If you would like to learn more about using R for statistical analysis, you might find the “Regression Models” course helpful.

B.2 Homework #2

Due date: Oct. 4 by midnight

For this assignment, you will submit the assignment to me by email by the due date. You should include three files in your submission:

  1. The final rendered Word document (rendered from an RMarkdown file).
  2. The original RMarkdown file used to create that final document.
  3. The dataset you used for the assignment.

For this assignment, start by picking a dataset either from your own research or something interesting available online (if you’re struggling to find something, check out Five Thirty Eight’s GitHub data repository). You will then use this dataset to practice what you’ve learned with R so far. This will also be a chance, if you’re using a dataset from your own research for me to get an idea of how you might be planning to use R in future research.

Very important note: Some research datasets have privacy constraints. This includes any datasets collected from human subjects, but can also include other datasets. For some projects, the principal investigator may prefer to keep the data private until publication of results. Before using a dataset for this project, please confirm with your research advisor that there are no constraints on the data. I do not plan to make the results of this assignment public, but I also do not want to us to be emailing back and forth a dataset with any constraints.

Using your dataset, create an RMarkdown document with the content listed below. Be sure to set the echo option to TRUE so all of your code will also print out to the Word document. Include in the RMarkdown document: (1) your full name; (2) the due date of the assignment; and (3) a title that includes “Homework 2.” These should all be set in RMarkdown, not changed in the Word output. Each section I’ve listed below should be included in the RMarkdown document as its own section, with a section header created using Markdown code.

  • Section 1: Description of the data: Describe the dataset you are using, both in terms of the content (what is this data measuring? how was it collected? what kinds of research questions are you hoping to use it to answer?) and in terms of its format (what type of file is it saved in? what if it is in a flat file, is it fixed width or delimited? if it is delimited, what is the delimiter? if it is binary, what is the program that would normally be used to open it?). (20 points)
  • Section 2: Reading the data into R: Include code that reads the data into R and assigns it to a dataframe object that you can use later in the document. Explain in the text which R function you used to read in the data (e.g., read_csv) and which package it came from (if it was not a base R function). If there were any special options you needed to use (e.g., skip to skip some rows without data), list those and explain why you used them. Next, include some code to clean the data (e.g., rename columns, convert any dates into a “Date” format). You can filter to certain rows if you would like, but do not filter out missing values, as we’ll want to learn more about those later. (20 points)
  • Section 3: Characteristics of the data: Describe the dataframe you just read in. How many rows does it have? How many columns? (Use inline code in the RMarkdown document to put this information into a sentence that reads “This dataframe has … rows and … columns.”) What are the names of the columns? What does each row measure (i.e., what is the unit of observation)? Include a table (using Markdown directly or kable) explaining what each column measures. This table should have three columns: (1) the column name in the R dataframe;
  1. a very brief description of what each column measures; and (3) the units (if any) of the measurement in the column. (20 points)
  • Section 4: Summary statistics: Pick three columns of the dataframe. Use the summarize function to get the following summaries of these columns: (1) minimum value; (2) maximum value; (3) mean value; (4) number of missing values. If there are missing values, make sure you use the appropriate options in summarizing these values to exclude those when calculating the minimum, maximum, and mean. Assign the result of this summarize call to a new R object, and print it out, so these summaries show up in your final, rendered Word document. (20 points)
  • Section 5: Visualizing the data: Create two plots of your dataframe. One should use a “statistical” geom (e.g., histogram, bar chart, boxplot) and one a “non-statistical” geom (e.g., scatterplot, line plot for time series). Explain why these plots help you learn more about this data and about the interesting research questions you’re hoping to explore with the data. Be sure to customize the final size of each plot in the Word document using the fig.width and fig.height commands. For each plot, also be sure to customize the x- and y-axis labels. Finally, explain how each plot is following at least two of the principles of “good graphics” covered in week 4 of the course (Chapter 4 of the book)—if necessary, use ggplot functions and options to make the plots comply with some of these principles. (20 points)

B.3 Homework #3

Due date: Oct. 20 by midnight

This homework includes several parts. Make sure you complete each part for full credit (100 points).

  1. Complete your homework in an RMarkdown document and knit it to either Word or PDF to submit. Change your global options so that no warnings, messages, or errors are printed out, but that so all code and results are printed. (10 points)
  2. Select one of the figures you created for the last homework. You will be improving that figure in this part of the homework.
    • Go through the six guidelines for good graphics we discussed in Chapter 6. Re-create the same plot you created in the last homework and, in a paragraph or two, list which of these elements you currently have in the plot and how these elements help make the plot more effective. (10 points)
    • Select three of these guidelines for which you either aren’t using them in the current graph or could add to what you currently have in the plot. Create a new version of the plot based on your choice. Include a paragraph explaining how these changes make the plot more effective and how you added them. (20 points)
  3. In the titanic package (available on CRAN), there’s a titanic_train dataset we’ll use for this question.
    • Load that dataset and create a dataframe object limited to the columns Survived and Age. Change the labeling for the Survived column so that it says “Survived” instead of “1” and “Died” instead of “0.” Print out the oldest five people who survived and who died (so, you’ll be printing out 10 rows total). (10 points)
    • Create a histogram of the ages of people in this dataset, faceted by whether or not they survived. Arrange these so that they are vertically aligned rather than side by side (Hint: check the nrow option when faceting). Change the label on the y-axis to be something that more precisely describes what’s being shown (rather than “count”) and add a title to the plot. (10 points)
    • Determine the mean age for each group (survived and died) for everyone who has a non-missing age, as well as the total number of people in the group and the number of people with age information missing for the group. Print out a two-row dataframe with this summary information (it should have columns for (1) which group (survived / died); (2) the mean age of people in the group with non-missing age data; (3) the total number of people in the group; and (4) the number of people in the group with missing age data). (15 points)
    • Run a statistical test of whether there is a difference in the mean ages between the two groups (Hint: you might want to google something like “test difference two means”). You can run this test either using a simple statistical test or by fitting a linear regression—you may take either approach. (15 points)
  4. There’s a radio show on National Public Radio called “Weekend Edition,” and each week they have a “Sunday Puzzle.” The puzzle for the week ending August 5, 2018 was this: “I said think of a familiar two-word phrase in eight letters with four letters in each word. The first word starts with M. And I said move the first letter of the second word to the end, and you get a regular eight-letter word, which, amazingly, other than the M, doesn’t share any sounds with the original two-word phrase. What phrase is it?” The winner had this to say about how he solved the problem: "I went to the National Puzzlers’ League website and pulled up a list of eight-letter words that start with the letter M and just started scanning through it…’. You can solve this puzzle using R (some of the stringr functions, like str_count and str_sub, will be particularly helpful). Write R code to solve this puzzle. (10 points)
    • Hint 1: Work backwords. Start with eight-letter words, and apply a backwards transform to the one described to transform them into two four-letter words.
    • Hint 2: This link might prove very useful: https://github.com/dwyl/english-words.

B.4 Homework #4

Due date: Nov. 3 by midnight

For this homework, you will use a data set collected by the Washington Post on homicides in 50 large U.S. cities. The data is available through their GitHub repository: https://github.com/washingtonpost/data-homicides. You can read their accompanying article at https://www.washingtonpost.com/graphics/2018/investigations/where-murders-go-unsolved/.

Submit your homework by email. You should include the Rmarkdown file (.Rmd) and the output Word document.

  1. Read in the data as an R object named homicidesand create a new column called city_name that combines the city and state like this “Baltimore, MD.” (15 points)
  2. Create a dataframe called unsolved with one row per city that gives the total number of homicides for the city and the number of unsolved homicides (those for which the disposition is “Closed without arrest” or “Open/No arrest”). (15 points)
  3. For the city of Baltimore, MD, use the prop.test function to estimate the proportion of homicides that are unsolved, as well as the 95% confidence interval for this proportion (we’re assuming that the data from the years covered by this dataset is a representative sample of Baltimore in a larger set of years). Print the output of the prop.test directly in your RMarkdown, and then save the output of prop.test as an R object and apply the tidy function from the broom package to this object and pull the estimated proportion and confidence intervals from the resulting tidy dataframe. (20 points)
  4. Now use what you learned from running prop.test for one city to run prop.test for all the cities. Your goal is to create the dataframe you need to create the figure shown below, where the points show the estimated proportions of unsolved homicides in each city and the horizontal lines show the estimated 95% confidence intervals. Do this all within a “tidy” pipeline, starting from the unsolved dataframe that you created for step 3. Use map2 from purrr to apply prop.test within each city and then map from purrr to apply tidy to this output. Use the unnest function from the tidyr package on the resulting list-column (from mapping tidy to the prop.test output list-column), with the option .drop = TRUE, to get your estimates back into a regular tidy data frame before plotting. (25 points)
  5. Create the plot shown below. Hint: Check out the geom_errorbarh geom with the height = 0 option to get the horizontal lines for the confidence intervals. To get full points, be very careful to make sure that all labeling, color choices, figure dimensions, etc., in your figure match the figure shown here. (25 points)

B.5 Homework #5

Due date: Nov. 17 by midnight

For this homework, you will continue using the dataset used for Homework #4. You will submit this project by sending me the link to a GitHub repository. For the write-up of this homework, you’ll need to write an RMarkdown document and render it to a pdf file within the R Project for the GitHub repository.

B.5.1 Setting up a GitHub repository for this project

Take the following steps to set up your GitHub repository for the homework (we will work on this as an in-course exercise):

  1. Set up a local (i.e., on your computer) R project with subdirectories for the data, writing (this is where you’ll put your Rmarkdown file and the output file), and figures.
  2. Download the Washington Post data and save it in an appropriate place in this R project directory.
  3. Initialize git for the R project and make your initial commit.
  4. Create an RMarkdown file for your homework answers in the appropriate subdirectory and save it. Commit this change with an appropriate commit message. (At this stage, you can look at your “History” in the git commit window to see the changes you’ve made so far.)
  5. Login to your GitHub account and create a new, blank repo with the same name as your R project. (If you have not already set up an SSH key to make it easier to push back and forth between your computer and GitHub, do so now.)
  6. On your computer, specify your GitHub repository as the remote for your project and do an initial push to send all the files to the GitHub repo. Look at your repository on GitHub to make sure it worked.
  7. Add a README.md file to the top level of your R project (on your local computer) using Markdown syntax and then commit the change locally and push to your GitHub repository.

For the README.md file, you can create a text file in RStudio and save it as “README.md.” You can then write this file using Markdown syntax (like RMarkdown, but without any code chunks).

As you work on your homework, make sure you commit regularly (with helpful commit messages). You should have at least 15 commit messages in your history for the repo by the time you turn in the homework.

Select one of the following two figures to create for the homework:

  1. Choice 1: Pick one city in the data. Create a map showing the locations of the homicides in that city, using the sf framework discussed in class. Use tigris to download boundaries for some sub-city geography (e.g., tracts, block groups, county subdivisions) to show as a layer underneath the points showing homicides. Use different facets for solved versus unsolved homicides and different colors to show the three race groups with the highest number of homicides for that city (you may find the fct_lump function from forcats useful for this).
  2. Choice 2: Recreate the graph shown below. It shows monthly homicides in Baltimore, with a reference added for the date of the arrest of Freddie Gray and color used to show colder months (November through April) versus warmer months (May through October). There is a smooth line added to help show seasonal and long-term trends in this data.

B.6 Homework #6

Due date: Wednesday, Dec. 8 by midnight

  1. Read the article Good Enough Practices in Scientific Computing by Wilson et al. (available here). In a half page, describe which of these “pretty good practices” you are incorporating for your group project. Also list one or two practices that you are not following but that would have made sense and how you could change your practices to follow them.
  1. Find an article in The R Journal that describes an R package that you could use in your own research or otherwise find interesting. Describe why the package was created and what you think it’s most interesting features are. In an R Markdown document, run one or two of the R examples included in the article.