Week 5: Regression

Conceptual Homework

You’ll be tempted to rush through some of the readings and exercises below. If you’ve had statistics before, this is pretty basic. However, it’s foundational for later work, so it’s worth another careful look.

Read the following chapters from FPP and do the following exercises (note, you only need to submit the exercises from FPP):

FPP, Chapter 7

• Do the assigned exercises.
• Review my notes on the x-y space and scatterplots from class. I recommend doing the exercises, but they are optional.

FPP, Chapter 8

• Do the assigned exercises.
• Review my notes on the correlation coefficient from class. Again, I recommend the exercises.

Optional: FPP, Chapter 9, and the assigned exercises

FPP, Chapter 12

• You might find it helpful to read chapters 10 and 11, time permitting.
• Do the assigned exercises.
• Review my notes on [notes on regression from class.

---

Computational Homework

Continuing from last week, you have a lot of flexibility on the computational assignments. Your task this week: learn something interesting from your data using the conceptual and computational tools we learned in class (i.e., scatterplots and regression).

You will do a simple data analysis and write a short paper summarizing your results. Feel free to use additional tools in your toolbox, such as histograms and data wrangling, or learn new tools as needed, but you must feature some of the tools we learned this week.

IMPORTANT: As you work through the steps below, feel free to copy files over from previous weeks’ assignments.

1. Prepare.
   1. Open RStudio and start a new RStudio Project called hw05 and initialize git.
   2. Open GitHub Desktop, add the local repo hw05-first-last, and publish the initial files up to GitHub as a part of the pos5737 organization.
   3. Create a data/, doc/, and R/, subdirectories of hw05.
   4. Save the raw version of your data set to data/.
2. Wrangle and analyze. In thoroughly commented R scripts, do the following:
   1. Wrangle the data into a clean, tidy data frame. You should have already done most or all of this work, but continue to make improvements as you see the opportunity or need.
   2. Create one or more scatterplots. Use geom_jitter() if your variables take on only a few values and you have a lot of overplotting. Spend some time making the figures look nice. Perhaps use facetting to see the scatterplots for different subsets of the data.
   3. Calculate one or more correlation coefficients. For these, you should use two variables that are roughly numeric.
   4. Fit one or more regression models. For these, you should use two variables that are roughly numeric.
3. Write a short paper explaining your results.. You may use LaTeX (recommended), Microsoft Word, or whatever you like to compose the document, but please include the PDF and keep it synced with the .tex or .docx source file. You should organize your work following the standard political science format, but please cover the following at some point.
   1. Write enough so that your reader understands how to interpret the key figure(s) and table(s).
   2. Spend some time explaining why someone should care about your point. (In my experience, this is both the hardest and most valuable part of a paper.)
   3. If you are able, explain how your paper fits with existing literature. There’s no need to read extra, but if you are aware of work on the topic, work to incorporate it.
4. Develop the README. Organize your README in a helpful way. You probably should include a summary of the basic argument, a link to the PDF of the manuscript, perhaps a figure or table or two in the README. When designing/updating your README, keep the goal in mind: to quickly bring the user up-to-speed with the project and help them understand and evaluate and extend your work.

When you are done, please make sure that you have completed all the steps above. Then submit your link to Canvas.