--- title: "Homework 2: Data Viz" author: "PUT YOUR NAME HERE" date: now date-format: "YYYY-MM-DDTHH:mm:ssZ" format: html: toc: true toc-depth: 2 embed-resources: true code-tools: true df-print: paged --- ```{r include = FALSE} # This chunk just sets up some styling (eg: default size of our images) knitr::opts_chunk$set( collapse = TRUE, warning = FALSE, message = FALSE, #error = TRUE, fig.height = 3, fig.width = 5, fig.align = 'center') ``` \ **DIRECTIONS** - Save this file as "homework_2" in your `DS112 > homework` folder. - Type your name in line 3 above (where it says "author"). - Type your responses in this template. You won't be able to Render it until after you've completed Exercise 0. - Do not modify the structure of this document (e.g. don't change section headers, spacing, etc). - Submit your rendered **HTML** file. We cannot grade qmd and pdf files. \ \ **GOALS** Practice some basic data exploration and data viz in guided settings. The content of the exercises is not necessarily in the order that we learned it, so you'll need to practice identifying appropriate tools for a given task. \ \ # Exercise 0 ## Part a We'll be using some data stored in the `mosaic` and `fivethirtyeight` packages. Follow the instructions from Activity 3 to: - Check whether you already have these packages. - If not, install them. ## Part b ```{r} # Load packages we'll need for HW 2: tidyverse, mosaic, fivethirtyeight ``` \ \ \ \ # Exercise 1 The `Birthdays` dataset in the `mosaic` package includes information on births in the U.S.. Run the chunk below which *wrangles* this information into a new dataset, `daily_births`, which records the total number of `births` per day in the U.S. between 1969 and 1988. **IMPORTANT: Use `daily_births`, not `Birthdays` in the exercises.** ```{r} # Create the daily_births dataset # Don't worry about the code. We'll learn this in the next unit! data("Birthdays") daily_births <- Birthdays %>% group_by(date) %>% summarize(births = sum(births)) %>% mutate(year = year(date), month = month(date, label = TRUE), day_of_month = mday(date), day_of_week = wday(date, label = TRUE)) ``` ## Part a Let's get to know the data! Complete the tasks indicated by each comment in the chunk below. (Put your code directly below the relevant comment.) ```{r} # Determine the number of data points we have # Check out the first 6 rows of the dataset # Check out the structure / class of each variable in the dataset (1 line of code) ``` \ \ ## Part b Notice from your work in part a that `day_of_week` and `month` are special types of *categorical* variables: `Ord.factors` or **ordinal factors**. Explain what makes `day_of_week` and `month` *ordinal*. (Looking up a definition of "ordinal" is fair, but you need to interpret the definition in the context of the variables here.) **ANSWER:** \ \ ## Part c What are the units of observation? **ANSWER:** \ \ \ \ # Exercise 2 Our general goal is to understand how births vary from day to day. Let's start by focusing on just the `births` variable. ## Part a Construct *two* plots of `births`. Each plot should use a different layer / geometric element. ```{r} # Plot 1 ``` ```{r} # Plot 2 ``` ## Part b Summarize your observations about daily births in the U.S. Tell a *cohesive* story here (not a list of distinct observations) that incorporates important aspects of quantitative variables: range, distribution / shape, typical outcome, outliers. **ANSWER:** \ \ \ \ # Exercise 3 Your plots above illustrate that there's something interesting happening here! Perhaps it can be explained by changes in birth rates over *time*. For example, we might wonder: Have births increased or decreased? Have there been any peaks or valleys in births? Are there any seasonal trends? We'll try to address these in the following parts. ## Part a Construct a plot of births over time. Only use 2 variables in your plot and make sure that every data point is represented in the plot (not just a summary of trends). We'll come back to this later. ```{r} ``` ## Part b Summarize your observation about births over time. Be sure to comment on the following: - What are the longer-term birth trends over time (on the scale of years)? - Are there any seasonal trends within years (on the scale of months)? - What other interesting or unusual patterns do you notice?!? **ANSWER:** \ \ \ \ # Exercise 4 Your above plot of births over time has an interesting pattern: there seem to be 2-3 distinct groups of points. Let's sleuth it out. ## Part a Starting from your plot of births over time from above, add a new aesthetic / variable that explains the distinction between the distinct groups. NOTE: Use your intuition (what variables do we have that might explain the 2 groups?) and play around. If you get stuck, move on to Exercise 5 and come back. ```{r} ``` ## Part b Explain in words *what* is going on here (e.g. "over time, there tend to be more births on... than on....") and *why* you think this is happening. NOTE: The "why" might be challenging, so just do your best. **ANSWER:** \ \ \ \ # Exercise 5 Let's dig into weekly and monthly birth trends. ## Part a Construct 2 plots that illustrate how birth rates might vary by month, one using density plots and another using boxplots. One of these will be more useful than the other! ```{r} ``` ## Part b Construct a plot that illustrates how birth rates might vary by day of the week. ```{r} ``` ## Part c Summarize your observations from parts a and b. Do birth rates tend to be higher in some months than others? Or higher on some days of the weeks than others? Explain. NOTE: If you skipped Exercise 4, you now have to hints. **ANSWER:** \ \ \ \ # Exercise 6 In the remaining exercises, we'll explore new data. The "Bechdel test" tests whether movies meet the following criteria: - there are at least 2 characters that identify as women - these characters talk to each other\ - at least 1 time, they talk about something other than men In the fivethirtyeight.com article ["The Dollar-And-Cents Case Against Hollywood's Exclusion of Women"](http://fivethirtyeight.com/features/the-dollar-and-cents-case-against-hollywoods-exclusion-of-women/), the authors analyze which Hollywood movies do/don't pass the test. Their data are available in the `fivethirtyeight` package\ Run the chunk below which *wrangles* this information into a new dataset, `new_bechdel`, which tidies up some of the categorical variables: ```{r} # Create the new_bechdel dataset # Don't worry about the code. We'll learn this in the next unit! data(bechdel) new_bechdel <- bechdel %>% mutate(clean_test = factor(clean_test, c("nowomen", "notalk", "men", "dubious", "ok"))) %>% mutate(half_decades = cut(year, breaks = seq(1969, 2014, by = 5))) ``` **IMPORTANT:** - Use `new_bechdel`, not `bechdel` in the exercises. - Before using the `new_bechdel` data, check out a codebook by typing `?bechdel` in the **CONSOLE**. ## Part a Construct a plot that addresses the following research question: Do bigger budgets (`budget_2013`) pay off with greater box office returns (`domgross_2013`)? In constructing this plot: - Include a layer that represents each individual film in the dataset. There are a lot of data points, so use transparency to improve the visualization. - Include a layer that highlights the *trends* in this relationship. - Pay attention to which of these is the *response* variable and which is the *predictor*. ```{r} ``` ## Part b Create a new plot, which tweaks the plot in part a, that addresses the following research question: Does the dependence of box office returns on budget differ among films that do and do not pass the Bechdel test? In examining the Bechdel test, use a variable that simply captures PASS or FAIL for each film. ```{r} ``` ## Part c Summarize your observations about the relationship between box office returns and budget, and how (if at all) this differs between films that do and do not pass the Bechdel test. **ANSWER:** \ \ \ \ # Exercise 7 The `clean_test` variable goes into more detail than the `binary` variable. Not only does it indicate whether a film passes the test, it indicates *why* a failing film fails. Read more about it in the codebook. ## Part a Construct a plot of the `clean_test` variable that allows us to explore the number of films that fall into each category. ```{r} ``` ## Part b Among films that fail the Bechdel test, what's the primary reason? **ANSWER:** \ ## Part c Consider two more questions: How are the films in the data set distributed across half decades, i.e. 5-year periods (`half_decades`)? And within each `half_decade`, how are the films distributed across Bechdel categories (`clean_test`)? Construct a single plot that answers both of these questions. Your y-axis should reflect *counts*, i.e. numbers of movies. ```{r} ``` \ \ \ \ # Exercise 8 NOTE: The following exercise is inspired by a similar exercise by Albert Kim, a `fivethirtyeight` package co-author. The plot in Exercise 7 illustrates that more of the films in our dataset are more recent, but it's not easy to examine the breakdown of the Bechdel test in each 5-year period. To that end, return to the fivethirtyeight.com article (linked above and in the codebook) and examine the plot titled "The Bechdel Test Over Time". ## Part a Summarize the trends captured by this plot. (How has the representation of women in movies evolved over time?) **ANSWER:** \ \ ## Part b Recreate this plot from the article!\ Your plot won't look *exactly* like the authors', but should be close to that below. You'll need to add the following layer to get a similar color scheme: ```{r eval=FALSE} scale_fill_manual(values = c("red", "salmon", "pink", "steelblue1", "steelblue4")) ``` ```{r, out.width = "800px",echo=FALSE, fig.alt='Proportional barplot of Bechdel test results across 5 year time periods. An increasing number of movies pass the Bechdel test over time but still only make up about 50% of the movies in the most recent time period.'} knitr::include_graphics("https://mac-stat.github.io/images/112/bechdel_hist.png") ``` ```{r fig.width = 8, fig.height = 5.5} ``` \ \ \ \ # Finalize your homework - Render your qmd one more time. Confirm that it appears as you expect it and that it's correctly formatted. If the formatting is amiss, we can't grade it :/ - If you're working on Mac's RStudio server, you have one more step that you should take at the end of each activity / assignment: export your files to your computer. To do so: - Go to the Files tab in the lower right pane. - Click the boxes next to the *two* homework files: homework_2.qmd and homework_2.html. - Still within the Files tab, click on the "More" button that has a gear symbol next to it. - Click "Export" then "Download". - The files were likely exported from the RStudio server to the Downloads folder on your computer. It's important to now move them to the "DS 112 \> Homework" folder that you created at the beginning of class. They are now there for safe keeping :) - Submit your **html** file to the Homework 2 assignment on Moodle. Do NOT submit a .qmd or pdf or any other file type -- we will not be able to grade them. - You're done with Homework 2. Congrats!!