worksheet_04.ipynb

Kernel: R

Worksheet 4: Effective Data Visualization

Lecture and Tutorial Learning Goals:

Expand your data visualization knowledge and tool set beyond what we have seen and practiced so far. We will move beyond scatter plots and learn other effective ways to visualize data, as well as some general rules of thumb to follow when creating visualizations. All visualization tasks this week will be applied to real world data sets.

After completing this week's lecture and tutorial work, you will be able to:

• Describe when to use the following kinds of visualizations:

  • scatter plots

  • line plots

  • bar plots

  • histogram plots

• Given a dataset and a question, select from the above plot types to create a visualization that best answers the question

• Given a visualization and a question, evaluate the effectiveness of the visualization and suggest improvements to better answer the question

• Identify rules of thumb for creating effective visualizations

• Define the three key aspects of ggplot objects:

  • aesthetic mappings

  • geometric objects

  • scales

• Use the ggplot2 library in R to create and refine the above visualizations using:

  • geometric objects: geom_point, geom_line, geom_histogram, geom_bar, geom_vline, geom_hline

  • scales: scale_x_continuous, scale_y_continuous

  • aesthetic mappings: x, y, fill, colour, shape

  • labelling: xlab, ylab, labs

  • font control and legend positioning: theme

  • flipping axes: coord_flip

  • subplots: facet_grid

• Describe the difference in raster and vector output formats

• Use ggsave to save visualizations in .png and .svg format

This worksheet covers parts of Chapter 4 of the online textbook. You should read this chapter before attempting the worksheet.

### Run this cell before continuing. 
library(tidyverse)
library(repr)
source("tests_worksheet_04.R")
source("cleanup_worksheet_04.R")

Question 0.1 True or False:
{points: 1}

For each of the items below, which are aesthetics in ggplot2?

A. colour
B. scale
C. x
D. fill
E. type

Assign either "true" or "false" for each to an object called answer0.1_L where L should be replaced by the letter corresponding to the question. For example, if above we had F) shape, we would set answer0.1_F <- "true".

# Make sure the correct answer is written in lower-case (true / false)
# Surround your answer with quotation marks.
# Replace the fail() with your answer. 

# your code here
fail() # No Answer - remove if you provide an answer

test_0.1()

Question 0.2 Multiple Choice:
{points: 1}

When deciding on the size of your visualization we recommend that you:

A. Only make the plot area (where the dots, lines, bars are) as big as needed

B. Make it as big as your screen allows

C. Use the default given by ggplot

Assign your answer to an object called answer0.2.

# Make sure the correct answer is written as an upper-case letter.
# Surround your answer with quotation marks.
# Replace the fail() with your answer. 

# your code here
fail() # No Answer - remove if you provide an answer

test_0.2()

Question 0.3 Multiple Choice:
{points: 1}

Under what circumstance would you use a 3D plot?

A. When you have 3 variables that you want to show the relationship between

B. When you want to emphasize the large difference between groups

C. When you need to grab attention of your audience

D. Never, we don't see well in 3D

Assign your answer to an object called answer0.3.

# Make sure the correct answer is an uppercase letter. 
# Surround your answer with quotation marks.
# Replace the fail() with your answer. 

# your code here
fail() # No Answer - remove if you provide an answer

test_0.3()

Question 0.4 Multiple Choice:
{points: 1}

What is the symbol used to add a new layer to a ggplot object?

A. <-

B. %>%

C. %&%

D. +

Assign your answer to an object called answer0.4.

# Make sure the correct answer is an uppercase letter. 
# Surround your answer with quotation marks.
# Replace the fail() with your answer. 

# your code here
fail() # No Answer - remove if you provide an answer

test_0.4()

Question 0.5 Item Arrangement
{points: 1}

Match the following geometric objects to their corresponding visualization purposes.

1. geom_histogram

2. geom_hline

3. geom_point

4. geom_line

5. geom_vline

6. geom_bar

A. Visualizing the distribution of data in one dimension
B. Visualizing the relationship between two variables in two dimensions
C. Comparing amounts
D. Denoting a quantity on the horizontal axis
E. Denoting a quantity on the vertical axis
F. Showing a trend with respect to an independent quantity

Assign each single character answer (e.g. "D") to an object called answer0.5_# where # should be replaced by the number corresponding to the item. For example, if we wanted to match 1 to F above, we would set `answer0.5_1 <- "F".

# Make sure each correct answer is a single uppercase letter. 
# Surround your answer with quotation marks.
# Replace the fail() with your answer. 

# your code here
fail() # No Answer - remove if you provide an answer

test_0.5()

1. World Vaccination Trends

Data scientists find work in all sectors of the economy and all types of organizations. Some work in collaboration with public sector organizations to solve problems that affect society, both at local and global scales. Today we will be looking at a global problem with real data from the World Health Organization (WHO). According to WHO, Polio is a disease that affects mostly children younger than 5 years old, and to date there is no cure. However, when given a vaccine, children can develop sufficient antibodies in their system to be immune to the disease. Another disease, Hepatitis B, is also known to affect infants but in a chronic manner. There is also a vaccine for Hepatitis B available.

The columns in the dataset are:

• who_region - The WHO region of the world

• yr - The year

• pct_vaccinated - Estimated percentage of people vaccinated in the region

• vaccine - Whether it's the polio or the hepatitis_b vaccine

We want to know two things. First, has there been a change in Polio or Hepatitis B vaccination patterns throughout the years? And if so, what is that pattern? Second, have the vaccination patterns for one of these dieseases changed more than the other? The goal for today is to produce a plot of the estimated percentage of people vaccinated per year. To do this, you will follow the steps outlined below.

The original datasets are available here:

• Polio: http://apps.who.int/gho/data/view.main.81605?lang=en

• Hepatitis B: http://apps.who.int/gho/data/view.main.81300?lang=en

The data set we will work with for this worksheet however is named world_vaccination.csv and it lives in the worksheet_04/data directory.

Question 1.0 Multiple Choice:
{points: 1}

Consider the columns/variables yr and pct_vaccinated? Are they:

A. both quantitative (e.g., numerical).

B. both categorical

C. one is categorical and one is quantitative

D. None of the above.

Assign your answer to an object called answer1.0.

# Make sure the correct answer is an uppercase letter. 
# Surround your answer with quotation marks.
# Replace the fail() with your answer. 

# your code here
fail() # No Answer - remove if you provide an answer

test_1.0()

Question 1.1
{points: 1}

Read the world_vaccination.csv file. Next, filter the data so that we don't have any NA's in our columns. We also want to filter out the (WHO) Global region as it is just the average of the other regions. Assign your filtered data to an object called world_vaccination.

When you want to filter for rows that aren't equal to something, in R you can use the != operator. For example to remove all the cars with 6 cylinders from the mtcars built-in R data set we would do the following:

head(mtcars)
filter(mtcars, cyl != 6)  %>% 
head()

#... <- read_...(...) %>%
# filter(!is.na(...), ... != ...)

# your code here
fail() # No Answer - remove if you provide an answer
head(world_vaccination)

test_1.1()

Question 1.2
{points: 1}

Create a scatter plot of the percentage of people vaccinated (y-axis) against year (x-axis) for all the regions in world_vaccination. Make sure to label your axes with human readable labels.

Assign your plot to an object called world_vacc_plot

# your code here
fail() # No Answer - remove if you provide an answer
world_vacc_plot

test_1.2()

Wow! That's a big plot! And unnecessarily so! It is so big that at 100% zoom, we cannot even read the labels and see all of the data within my screen! Given that the "ink" (dots) in this plot are not very complex, let's take charge and reduce the plot size so that the plot area where the ink is, is only as big as it needs to be. We do this using the options function from the repr R package. In past worksheets and tutorials we did this for you, from now on you need to do this yourself and choose an approprate plot size!

The options function takes 2 arguments for plot size, repr.plot.width and repr.plot.height. Uncomment the first line of code in the code cell below and fill in the ... to choose a more approprate size and and replot the figure afterwards below.

Note: if you get a weird error saying unable to start device 'png', it's because you set your plot's width and height too large - try a smaller number for each. If you get an error saying invalid 'width' / 'height' argument, it may be because you set the width/height too small - try a larger number.

#options(repr.plot.width = ..., repr.plot.height = ...) 
world_vacc_plot

Question 1.3 Multiple Choice:
{points: 1}

Now that we see how the percentage of people vaccinated with each of these vaccines varies over time, we should now start to look if there is a difference between the percentage vaccinated between the two different diseases. What should we do next to compare the differences (if they exist) most effectively?

A. Filter the data by the type of vaccine and make two separate plots

B. Colour the data by the type of vaccine

C. Have a different shaped "dot"/point for each type of vaccine

D. Colour the data by the type of vaccine, and have a different shaped "dot"/point for each type of vaccine

Assign your answer to an object called answer1.3.

# Make sure the correct answer is an uppercase letter. 
# Surround your answer with quotation marks.
# Replace the fail() with your answer. 

# your code here
fail() # No Answer - remove if you provide an answer

test_1.3()

Question 1.4
{points: 1}

Now that we know how we will separate the data for our visualization, let's do it. Start by copying your code from Question 1.2.

Next, add an aes function inside of the geom_point function to map vaccine to colour and shape. So your geom_point function and layer should look something like this:

geom_point(aes(colour = ..., shape = ...))  +

Finally, make sure you change all the axes and legends so they have nicely formatted, human-readable labels. To do this, we will remove the xlab and ylab label functions and just use the more general labs function, specifying all the aesthetics we want to label:

labs(x = "...", y = "...", colour = "...", shape = "...")

Assign your answer to an object called compare_vacc_plot.

# your code here
fail() # No Answer - remove if you provide an answer
compare_vacc_plot

test_1.4()

Now we see that although the dates where the percentage vaccinated became > 0 for each vaccine type started out in different years, they both increased at similar rates and are currently resting at about the same amount of percentage vaccinated. There is some variation that still exists in the data however, and perhaps that could be attributed to region? Let's create some more visualizations to see if that is indeed the case!

To get started, let's focus on the Polio vaccine data, and then we'll look at both together.

Question 1.5
{points: 1}

Create a data frame object named polio that contains only the rows where the vaccine is "polio":

# your code here
fail() # No Answer - remove if you provide an answer
head(polio)

test_1.5()

Question 1.6
{points: 1}

Now create a scatter plot using the polio data where percentage vaccinated is on the y-axis, year is on the x-axis and each group has a different coloured point, and a different shape. Name it polio_regions. You might want to use options to change the plot size if the size of the last few plots isn't ideal for this plot.

#polio_regions <- ...

# your code here
fail() # No Answer - remove if you provide an answer
polio_regions

test_1.6()

Question 1.7.1
{points: 1}

Although when we have multiple groups, its easier for us to see the differences when we change point colour and shape, at some point there are too many groups to keep things straight. We are approaching that on the plot above, and so we need to do something different... One thing we could try is to change the point to a line to reduce the noise/chaos of the plot above. We would also not have a shape. Do that in the cell below and name the plot object polio_regions_line.

#polio_regions_line <- ...

# your code here
fail() # No Answer - remove if you provide an answer
polio_regions_line

test_1.7.1()

Question 1.7.2
{points: 1}

One thing that is still not ideal with the visualization above is the legend title is not very readable. Let's add another layer to polio_regions_line to do that. To do this we use the labs function and choose the aesthetic mapping (here color) that we want to apply the legend title to. Also, given that we created an object from our previous plot, we do not need to retype all our code, but instead can just say:

[your plot object] <- [your plot object] +
                        [new layer code]

#polio_regions_line <- polio_regions_line +
#    labs(... = "Region of the world")

# your code here
fail() # No Answer - remove if you provide an answer
polio_regions_line

test_1.7.2()

Question 1.8
{points: 1}

Now that we know how to effectively plot the percentage vaccinated against Polio over time for each region, how might we compare this to what we see for each region for the percentage vaccinated against Hepatitis B? In this case we would like two side-by-side or two vertically arranged plots. If that data are in the same data frame (as ours were in the world_vaccination data frame) then we can use a technique called facetting to do this. We saw facetting last week, and now we will take some time to learn it.

There are two facetting functions in R, the one we will see here is facet_grid. The basic syntax for this ggplot layer is the following:

# creates side by side plots for each member of the category in COLUMN_X
facet_grid(. ~ COLUMN_X)

or

# creates vertically arranged plots for each member of the category in COLUMN_X
facet_grid(COLUMN_X ~ .)

Create a plot like the one named polio_regions_line but instead of using the polio data frame, use the world_vaccination data frame, and facet on the column vaccine so that the two plots are side-by-side. Name this plot object side_by_side_world. Make sure you set options to make the plot size fit well within the worksheet.

#... <- ... +
#    geom_...(...) +
#    labs(...) +
#    xlab(...) + 
#    ylab(...) +
#    facet_grid(...)

# your code here
fail() # No Answer - remove if you provide an answer
side_by_side_world

test_1.8()

Question 1.9.1
{points: 1}

Now use facet_grid to arrange the same two plots vertically. Name this plot vertical_world. Again, make sure you set options to create a suitable plot size.

# your code here
fail() # No Answer - remove if you provide an answer
vertical_world

test_1.9.1()

Which arrangement is better? Depends on what you are asking! If you are interested in comparing the rate at which things changed over time, then the vertical arrangement is more effective. However, if you are interested in comparing the exact percentage values between the lines at certain points then the side-by-side arrangement is more effective.

Question 1.9.2 Multiple Choice:
{points: 1}

Which WHO region had the greatest progress in the shortest period of time in either Hepatitis B and in Polio (using the data we plotted above)?

A. Americas

B. Eastern Mediterranean

C. Europe

D. Western Pacific

Assign your answer to an object called answer1.9.2.

# Make sure the correct answer is an uppercase letter. 
# Surround your answer with quotation marks.
# Replace the fail() with your answer. 

# your code here
fail() # No Answer - remove if you provide an answer

test_1.9.2()

2. Fast-Food Chains in the United States

With their cheap meals and convenient drive-thrus, fast food restaurants are a growing demand in many countries. Despite their questionable ingredients and nutritional value, most Americans count on fast food in their daily lives (they are often delicious and so hard to resist...).

Source: https://media.giphy.com/media/NS6SKs3Lt8cPHhe0es/giphy.gif

According to Wikipedia,

Fast food was originally created as a commercial strategy to accommodate the larger numbers of busy commuters, travelers and wage workers who often didn't have the time to sit down at a public house or diner and wait the normal way for their food to be cooked. By making speed of service the priority, this ensured that customers with strictly limited time (a commuter stopping to procure dinner to bring home to their family, for example, or an hourly laborer on a short lunch break) were not inconvenienced by waiting for their food to be cooked on-the-spot (as is expected from a traditional "sit down" restaurant). For those with no time to spare, fast food became a multi-billion dollar industry.

Currently, fast food is the norm and lots of businesses are investing in advertisement as well as new ideas to make their chain stand out in the sea of restaurants. In fact, one business is hiring you. They want to know the layout of the landscape:

1. Which is the most popular franchise on the west coast of the US?

2. Which state has the highest number of fast-food restuarants?

3. Is the most dominant franchise consistent across the west coast?

In this assignment, you will pretend to assist in the opening of a new restaurant somewhere on the west coast of the United States (California, Oregon, or Washington). Your goal is to figure out which chain to recommend and figure out which state would be the least competitive.

The columns in the dataset are:

• name - The name of the restaurant

• st - State

Question 2.1 Multiple Choice:
{points: 1}

From the list below, what are you not trying to determine:

A. The west coast frachise with the greatest popularity

B. The least dominant franchise consistent across the west coast

C. The state on the west coast with the greatest number of fast-food restuarants

Assign your answer to an object called answer 2.1.

# Make sure the correct answer is an uppercase letter. 
# Surround your answer with quotation marks.
# Replace the fail() with your answer. 

# your code here
fail() # No Answer - remove if you provide an answer

test_2.1()

Question 2.2
{points: 1}

Read the fast_food.csv file (found in the worksheet_04/data directory) and assign it to an object called fast_food.

# your code here
fail() # No Answer - remove if you provide an answer
head(fast_food)

test_2.2()

Question 2.3
{points: 1}

Next, find the top 9 restaurants (in terms of number of locations) on the west coast (in the states "CA", "WA" or "OR") and name them top_restaurants

Fill in the ... in the cell below, removing the fail().

Assign your answer to an object called top_restaurants.

#top_restaurants <- fast_food %>%
#    filter(st %in% c("CA", "WA", "OR")) %>% 
#    group_by(...) %>%
#    ...(n = n()) %>%
#    ...(...) %>% 
#    ...(...)

# your code here
fail() # No Answer - remove if you provide an answer
top_restaurants

test_2.3()

Question 2.4
{points: 1}

Plot the counts for the top 9 fast food restaurants on the west coast as a bar chart using geom_bar. The number of restaurants should be on the y-axis and the restaurant names should be on the x-axis. Because we are not counting up the number of rows in our data frame, but instead are plotting the actual values in the n column, we need to use the stat = "identity" argument inside geom_bar.

To do this fill in the ... in the cell below. Copy and paste your finished answer into the fail(). Make sure to label your axes and choose an appropriate figure size using options.

Assign your answer to an object called count_bar_chart.

#... <- ...(..., aes(x = ..., y = ...)) + 
#    geom_bar(...) +
#    xlab(...) +
#    ylab(...)

# your code here
fail() # No Answer - remove if you provide an answer
count_bar_chart

test_2.4()

Question 2.5
{points: 1}

The x-axes labels don't look great unless you make the bars on the bar plot above quite wide, wider than are actually useful or effective. What can we do? There are two good solutions to this problem.

Part A: We can add a theme layer and rotate the labels. Choose an angle that you think is appropriate. Choose something between 20 and 90 for the angle argument. Use the hjust = 1 argument to ensure your labels don't sit on top of the bars as you rotate them (try removing that argument and see what happens...).

Name the resulting plot count_bar_chart_A.

Part B: We can also simply use horizontal bars. Use the coord_flip function to achieve this effect.

Name the resulting plot count_bar_chart_B.

#PART A
#count_bar_chart_A <- ... + 
#    theme(axis.text.x = element_text(angle = ..., hjust = 1))

# your code here
fail() # No Answer - remove if you provide an answer
count_bar_chart_A

#PART B
#count_bar_chart_B <- ... + 
#    coord_flip()

# your code here
fail() # No Answer - remove if you provide an answer
count_bar_chart_B

test_2.5_A()
test_2.5_B()

Question 2.6
{points: 1}

Which is the most popular franchise on the west coast? Save your answer as answer2.6 and be sure to surround the restuarant name with quotations. Pay attention to case and punctuation when answering.

# your code here
fail() # No Answer - remove if you provide an answer

test_2.6()

Question 2.7
{points: 1}

Next, let's find which state on the west coast has the greatest number of those top 9 fast-food restuarants we found earlier. To do this we need to use the names in top_restaurants to get the counts of each restaurant in each of the 3 states from the fast_food data frame. You will need to use semi_join to do this (to get the intersection of two data frames). Name this data frame state_counts. Fill in the ... in the cell below. Copy and paste your finished answer into the fail().

If you are interested in learning more about joining data frames in R, see this cheatsheet.

#... <- fast_food %>%
#    semi_join(top_restaurants) %>% # semi_join gives the intersection of two data frames
#    filter(st %in% c("CA", "WA", "OR")) %>% 
#    ...(...) %>%
#    ...(n = n()) 

# your code here
fail() # No Answer - remove if you provide an answer
head(state_counts)

test_2.7()

Question 2.8
{points: 1}

Now, create a bar plot that has restaurant count on the y-axis and US state on the x-axis. Name the plot state_counts_plot. Remember to choose an appropriate plot size using options.

#state_counts_plot <- ...

# your code here
fail() # No Answer - remove if you provide an answer
state_counts_plot

test_2.8()

Question 2.9.1
{points: 1}

Which state (CA, OR, WA) has the highest number of fast-food restuarants? Save your answer as answer2.9.1 and be sure to surround the state initialism with quotations (e.g. "NY"). Pay attention to case and punctuation when answering.

# Surround your answer with quotation marks.
# Replace the fail() with your answer. 

# your code here
fail() # No Answer - remove if you provide an answer

test_2.9.1()

Consider the populations of California (325.7 million), Oregon (4.143 million) and Washington (7.406 million) (source: United States Census Bureau). Discuss with your neighbour about whether using the raw restaurant count for each state is the best measure of competition? Would restaurant per capita be a better alternative?

Question 2.9.2
{points: 1}

Is the most dominant/top franchise consistent across the west coast? To answer this question we need a data frame that has three columns: name (restaurant), st and n (restaurant count). We will need to use the semi-join strategy as we did above to use the names in top_restaurants to get the counts of each restaurant in each of the 3 states from the fast_food data frame. This time, however, we'll need to group_by both name and st. Name this new data frame top_n_state. We only want to look at those top 9 fast-food restuarants we found earlier.

#... <- fast_food %>%
#    semi_join(top_restaurants) %>% # semi_join gives the intersection of two data frames
#    filter(st %in% c("CA", "WA", "OR")) %>% 
#    ...(..., ...) %>%
#    ...(...) 

# your code here
fail() # No Answer - remove if you provide an answer
head(top_n_state)

test_2.9.2()

Question 2.9.3
{points: 1}

Plot the counts (y-axis) for the top 9 fast food restaurants (x-axis) on the west coast, per US State (group), as a bar chart using geom_bar. Use fill = name inside aes to color the restaurant's by name. Use position = "dodge" inside geom_bar to group the bars by state. To rename the legend, use a labs layer. This time within labs use the fill argument instead of color (this is because you need to modify the asthetic that the legend was made from, here it was fill, not color as earlier in the worksheet).

To do this fill in the ... in the cell below. Copy and paste your finished answer into the fail(). Make sure to label your axes and choose an appropriate plot size.

Assign your answer to an object called top_n_state_plot.

#... <- ggplot(..., aes(x = st, y = n, fill = ...)) + 
#    ...(stat = ..., position = "...") + 
#   xlab(...) +
#   ylab(...) +
#   labs(fill = "Restaurant") 

# your code here
fail() # No Answer - remove if you provide an answer
top_n_state_plot

test_2.9.3()

How easy is that to compare the restaurants and states to answer our question: Is the most dominant/top franchise consistent across the West-Coast? If we carefully look at this plot we can pick answer this question, but it takes us a while to process this. If we instead visualize this as a stacked bar chart using proportions instead of counts we might be able to do this easier (making it a more effective visualization).

Question 2.9.4
{points: 1}

Copy your code from Question 2.9.2 and modify position = "dodge" to position = "fill" to change from doing a grouped bar chart to a stacked bar chart with the data represented as proportions instead of counts.

#top_n_state_plot <- ...

# your code here
fail() # No Answer - remove if you provide an answer
top_n_state_plot

test_2.9.4()

Question 2.9.5
{points: 1}

Is the most dominant franchise consistent across the west coast? Answer "yes" or "no". Save your answer as answer2.9.5 and be sure to surround the answer with quotations. Pay attention to case and punctuation when answering.

# Make sure the correct answer is written in lower-case (yes / no)
# Surround your answer with quotation marks.
# Replace the fail() with your answer. 

# your code here
fail() # No Answer - remove if you provide an answer

test_2.9.5()

We are just scratching the surface of how to create effective visualizations in R. For example, we haven't covered how to change from the default colors palette ggplot2 provides. We'll learn more in the tutorial, but it's a big world out there; to learn more, visit the links in the reading and practice, practice, practice! Go forth and make beautiful and effective plots!