introduction
prerequisites
explore
carpentry
design
report
exercises
references
“By encoding the data as positions along two orthogonal scales, a two-dimensional scatter plot reveals the strength of the relation between two variables, together with outliers. It can easily identify subsets with different markers, although these must be well contrasted, lest the display look cluttered.” (Doumont, 2009)
Data characteristics
- Two quantitative variables
- One or more categorical variables (optional)
- A key variable if data are not coordinatized
Graph characteristics
- Two-dimensional orthogonal axes with scales that increase from left to right and from top to bottom
- The levels of a category can be distinguished by data marker type or color or with facets
- Facets, if used, have identical scales
D3 Correlations data and graph requirements
Project setup
- Start every work session by launching the RStudio Project file for
the course, e.g.,
portfolio.Rproj - Ensure your project directory structure satisfies the course requirements
Ensure you have installed the following packages. See install packages for instructions if needed.
- tidyverse: The ‘tidyverse’ is a set of packages that work in harmony because they share common data representations and ‘API’ design. This package is designed to make it easy to install and load multiple ‘tidyverse’ packages in a single step. Learn more about the ‘tidyverse’ at https://tidyverse.org.
- graphclassmate: An R package with companion materials for a course in data visualization. The package provides data sets structured for a variety of graph types plus a ggplot2 theme.
- VGAMdata: Data sets to accompany the VGAM package and the book “Vector Generalized Linear and Additive Models: With an Implementation in R” (Yee, 2015) DOI:10.1007/978-1-4939-2818-7. These are used to illustrate vector generalized linear and additive models (VGLMs/VGAMs), and associated models (Reduced-Rank VGLMs, Quadratic RR-VGLMs, Row-Column Interaction Models, and constrained and unconstrained ordination models in ecology).
Scripts to initialize
explore/ 0503-scatterplot-oly12-explore.R
carpentry/ 0503-scatterplot-oly12-data.R
design/ 0503-scatterplot-oly12.R
And start each file with a minimal header
# your name
# date
# load packages
library("tidyverse")Duplicate the lines of code in the session one chunk at a time. Save, Source, and compare your results to the results shown.
Open the explore script you initialized earlier. Load the package that has the data.
library("VGAMdata")Data about athletes in the 2012 summer Olympics is in VGAMdata::oly12.
The number of observations is greater than the number of names, probably
because some athletes compete in more than one event.
prep <- oly12
str(prep)
#> 'data.frame': 10384 obs. of 14 variables:
#> $ Name : Factor w/ 10366 levels "Aaron Brown",..: 5353 121 4117 16 6033 5686 6061 6765 2738 3854 ...
#> $ Country: Factor w/ 205 levels "Afghanistan",..: 144 195 68 125 154 68 8 125 94 3 ...
#> $ Age : int 23 33 30 24 26 27 30 23 27 19 ...
#> $ Height : num 1.7 1.93 1.87 NA 1.78 1.82 1.82 1.87 1.9 1.7 ...
#> $ Weight : int 60 125 76 NA 85 80 73 75 80 NA ...
#> $ Sex : Factor w/ 2 levels "F","M": 2 2 2 2 1 2 1 2 2 2 ...
#> $ DOB : Date, format: "1989-02-06" NA ...
#> $ PlaceOB: Factor w/ 4108 levels "","Aachen (GER)",..: 2486 3302 398 48 3436 1 1 1 1172 2266 ...
#> $ Gold : int 0 0 0 0 0 0 0 0 0 0 ...
#> $ Silver : int 0 0 0 0 0 0 0 0 0 0 ...
#> $ Bronze : int 0 0 0 0 0 0 0 0 0 0 ...
#> $ Total : int 0 0 0 0 0 0 0 0 0 0 ...
#> $ Sport : Factor w/ 42 levels "Archery","Athletics",..: 27 2 2 7 2 25 29 22 2 7 ...
#> $ Event : Factor w/ 763 levels "Group All-Around",..: 350 405 251 443 699 406 726 403 248 491 ...It is a data frame, so I’ll convert it to a tibble
prep <- prep %>%
as_tibble() %>%
glimpse()
#> Observations: 10,384
#> Variables: 14
#> $ Name <fct> Lamusi A, A G Kruger, Jamale Aarrass, Abdelhak Aatakni...
#> $ Country <fct> People's Republic of China, United States of America, ...
#> $ Age <int> 23, 33, 30, 24, 26, 27, 30, 23, 27, 19, 37, 28, 28, 28...
#> $ Height <dbl> 1.70, 1.93, 1.87, NA, 1.78, 1.82, 1.82, 1.87, 1.90, 1....
#> $ Weight <int> 60, 125, 76, NA, 85, 80, 73, 75, 80, NA, NA, NA, 60, 6...
#> $ Sex <fct> M, M, M, M, F, M, F, M, M, M, M, M, F, F, M, F, M, M, ...
#> $ DOB <date> 1989-02-06, NA, NA, 1988-09-02, NA, 1984-06-09, NA, 1...
#> $ PlaceOB <fct> NEIMONGGOL (CHN), Sheldon (USA), BEZONS (FRA), AIN SEB...
#> $ Gold <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, ...
#> $ Silver <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, ...
#> $ Bronze <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, ...
#> $ Total <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, ...
#> $ Sport <fct> Judo, Athletics, Athletics, Boxing, Athletics, Handbal...
#> $ Event <fct> "Men's -60kg", "Men's Hammer Throw", "Men's 1500m", "M...Examine the quantitative variables. I’m not interested in medals, so age, height, and weight are potentially useful data to graph.
prep_numeric <- prep %>%
select_if(is.numeric) %>%
glimpse()
#> Observations: 10,384
#> Variables: 7
#> $ Age <int> 23, 33, 30, 24, 26, 27, 30, 23, 27, 19, 37, 28, 28, 28,...
#> $ Height <dbl> 1.70, 1.93, 1.87, NA, 1.78, 1.82, 1.82, 1.87, 1.90, 1.7...
#> $ Weight <int> 60, 125, 76, NA, 85, 80, 73, 75, 80, NA, NA, NA, 60, 64...
#> $ Gold <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0...
#> $ Silver <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0...
#> $ Bronze <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0...
#> $ Total <int> 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0...Keep age, height, and weight, and examine the quantiles. Results show we have some NAs in height and weight.
# keep age, height, and weight
prep_numeric <- prep_numeric %>%
select(Age, Height, Weight)
summary(prep_numeric)
#> Age Height Weight
#> Min. :13.00 Min. :1.320 Min. : 36.00
#> 1st Qu.:22.00 1st Qu.:1.690 1st Qu.: 61.00
#> Median :25.00 Median :1.770 Median : 70.00
#> Mean :26.07 Mean :1.769 Mean : 72.85
#> 3rd Qu.:29.00 3rd Qu.:1.850 3rd Qu.: 81.00
#> Max. :71.00 Max. :2.210 Max. :218.00
#> NA's :561 NA's :1280Let’s see how many levels of each factor we have,
prep_factor <- prep %>%
select_if(is.factor) %>%
glimpse()
#> Observations: 10,384
#> Variables: 6
#> $ Name <fct> Lamusi A, A G Kruger, Jamale Aarrass, Abdelhak Aatakni...
#> $ Country <fct> People's Republic of China, United States of America, ...
#> $ Sex <fct> M, M, M, M, F, M, F, M, M, M, M, M, F, F, M, F, M, M, ...
#> $ PlaceOB <fct> NEIMONGGOL (CHN), Sheldon (USA), BEZONS (FRA), AIN SEB...
#> $ Sport <fct> Judo, Athletics, Athletics, Boxing, Athletics, Handbal...
#> $ Event <fct> "Men's -60kg", "Men's Hammer Throw", "Men's 1500m", "M...
sapply(prep_factor, nlevels)
#> Name Country Sex PlaceOB Sport Event
#> 10366 205 2 4108 42 763Based on the number of levels present, we might want to condition on sex, sport, or event. So the first thing I’ll do is select the columns just mentioned and filter for complete rows.
prep <- prep %>%
select(Name, Country, Age, Height, Weight, Sex, Sport, Event) %>%
filter(complete.cases(.)) %>%
glimpse()
#> Observations: 9,038
#> Variables: 8
#> $ Name <fct> Lamusi A, A G Kruger, Jamale Aarrass, Maria Abakumova,...
#> $ Country <fct> People's Republic of China, United States of America, ...
#> $ Age <int> 23, 33, 30, 26, 27, 30, 23, 27, 28, 28, 22, 22, 25, 25...
#> $ Height <dbl> 1.70, 1.93, 1.87, 1.78, 1.82, 1.82, 1.87, 1.90, 1.73, ...
#> $ Weight <int> 60, 125, 76, 85, 80, 73, 75, 80, 60, 64, 62, 63, 69, 8...
#> $ Sex <fct> M, M, M, F, M, F, M, M, F, F, M, M, M, M, F, M, M, F, ...
#> $ Sport <fct> Judo, Athletics, Athletics, Athletics, Handball, Rowin...
#> $ Event <fct> "Men's -60kg", "Men's Hammer Throw", "Men's 1500m", "W...Remove duplicate rows using dplyr::distinct(). The glimpse shows the
same number of rows as before the operation so we conclude there were no
duplicates.
prep <- prep %>%
distinct() %>%
glimpse()
#> Observations: 9,038
#> Variables: 8
#> $ Name <fct> Lamusi A, A G Kruger, Jamale Aarrass, Maria Abakumova,...
#> $ Country <fct> People's Republic of China, United States of America, ...
#> $ Age <int> 23, 33, 30, 26, 27, 30, 23, 27, 28, 28, 22, 22, 25, 25...
#> $ Height <dbl> 1.70, 1.93, 1.87, 1.78, 1.82, 1.82, 1.87, 1.90, 1.73, ...
#> $ Weight <int> 60, 125, 76, 85, 80, 73, 75, 80, 60, 64, 62, 63, 69, 8...
#> $ Sex <fct> M, M, M, F, M, F, M, M, F, F, M, M, M, M, F, M, M, F, ...
#> $ Sport <fct> Judo, Athletics, Athletics, Athletics, Handball, Rowin...
#> $ Event <fct> "Men's -60kg", "Men's Hammer Throw", "Men's 1500m", "W...With only three quantitative variables to study, we might expect a correlation between height and weight but not age. Let’s graph that.
ggplot(data = prep, aes(x = Height, y = Weight)) +
geom_jitter()
Looks like a possible correlation. Next, condition on sex,
ggplot(data = prep, aes(x = Height, y = Weight, color = Sex)) +
geom_jitter()
And a facet by sport,
ggplot(data = prep, aes(x = Height, y = Weight, color = Sex)) +
geom_jitter() +
facet_wrap(vars(Sport))
We have a number of sports with only a few values. let’s count,
n_athletes <- prep %>%
count(Sport) %>%
arrange(desc(n))
kable(n_athletes)| Sport | n |
|---|---|
| Athletics | 1900 |
| Swimming | 859 |
| Football | 589 |
| Rowing | 512 |
| Hockey | 383 |
| Sailing | 351 |
| Judo | 341 |
| Shooting | 338 |
| Handball | 304 |
| Wrestling | 296 |
| Volleyball | 271 |
| Basketball | 258 |
| Water Polo | 249 |
| Fencing | 229 |
| Weightlifting | 229 |
| Canoe Sprint | 224 |
| Equestrian | 181 |
| Cycling - Road | 176 |
| Cycling - Track | 160 |
| Badminton | 159 |
| Tennis | 156 |
| Table Tennis | 151 |
| Diving | 128 |
| Archery | 114 |
| Triathlon | 103 |
| Beach Volleyball | 93 |
| Canoe Slalom | 80 |
| Cycling - Mountain Bike | 70 |
| Modern Pentathlon | 66 |
| Cycling - BMX | 42 |
| Cycling - Road, Cycling - Track | 16 |
| Cycling - Mountain Bike, Cycling - Road, Cycling - Track | 3 |
| Cycling - Mountain Bike, Cycling - Track | 3 |
| Cycling - Mountain Bike, Cycling - Road | 2 |
| Athletics, Triathlon | 1 |
| Cycling - BMX, Cycling - Track | 1 |
The largest level of Sport is “Athletics” which includes all the competitive running, jumping, throwing, and walking. In contrast, there are a few with less than 50 athletes. I’ll drop those Sports for the data display.
n_small <- n_athletes %>%
filter(n < 50)
drop_these <- n_small$Sport %>%
droplevels() %>%
as.character()
prep <- prep %>%
filter(!Sport %in% drop_these)
ggplot(data = prep, aes(x = Height, y = Weight, color = Sex)) +
geom_jitter() +
facet_wrap(vars(Sport))
There is one weight outlier in Judo I think I’ll drop because he makes all the scales span more than need be for the rest of the facets.
prep <- prep %>%
filter(Weight < 200)
ggplot(data = prep, aes(x = Height, y = Weight, color = Sex)) +
geom_jitter() +
facet_wrap(vars(Sport))
Later, in design, we’ll continue to edit the visual elements, but this seems to be an acceptable layout.
Open the carpentry script you initialized earlier.
A data carpentry file typically begins by reading the source data file. In this case of course, the data are provided with a package.
library("tidyverse")
data(oly12, package = "VGAMdata")I would like to take another look at the sports levels to see if some should be recoded.
levels(oly12$Sport)
#> [1] "Archery"
#> [2] "Athletics"
#> [3] "Athletics, Triathlon"
#> [4] "Badminton"
#> [5] "Basketball"
#> [6] "Beach Volleyball"
#> [7] "Boxing"
#> [8] "Canoe Slalom"
#> [9] "Canoe Sprint"
#> [10] "Cycling - BMX"
#> [11] "Cycling - BMX, Cycling - Track"
#> [12] "Cycling - Mountain Bike"
#> [13] "Cycling - Mountain Bike, Cycling - Road"
#> [14] "Cycling - Mountain Bike, Cycling - Road, Cycling - Track"
#> [15] "Cycling - Mountain Bike, Cycling - Track"
#> [16] "Cycling - Road"
#> [17] "Cycling - Road, Cycling - Track"
#> [18] "Cycling - Track"
#> [19] "Diving"
#> [20] "Equestrian"
#> [21] "Fencing"
#> [22] "Football"
#> [23] "Gymnastics - Artistic"
#> [24] "Gymnastics - Rhythmic"
#> [25] "Handball"
#> [26] "Hockey"
#> [27] "Judo"
#> [28] "Modern Pentathlon"
#> [29] "Rowing"
#> [30] "Sailing"
#> [31] "Shooting"
#> [32] "Swimming"
#> [33] "Synchronised Swimming"
#> [34] "Table Tennis"
#> [35] "Taekwondo"
#> [36] "Tennis"
#> [37] "Trampoline"
#> [38] "Triathlon"
#> [39] "Volleyball"
#> [40] "Water Polo"
#> [41] "Weightlifting"
#> [42] "Wrestling"I am going to recode some of the cycling levels to combine them and drop unused levels.
oly12 <- oly12 %>%
mutate(Sport = fct_recode(Sport,
"BMX" = "Cycling - BMX",
"BMX" = "Cycling - BMX, Cycling - Track",
"Mountain Bike" = "Cycling - Mountain Bike",
"Mountain Bike" = "Cycling - Mountain Bike, Cycling - Road",
"Mountain Bike" = "Cycling - Mountain Bike, Cycling - Road, Cycling - Track",
"Mountain Bike" = "Cycling - Mountain Bike, Cycling - Track",
"Road Cycling" = "Cycling - Road",
"Road Cycling" = "Cycling - Road, Cycling - Track",
"Pentathlon" = "Modern Pentathlon"
)) %>%
droplevels() Now I’ll collect all the steps from the explore file that we intend to keep.
# sports to drop because of limited participation
drop_these <- oly12 %>%
count(Sport) %>%
filter(n < 50) %>%
select(Sport) %>%
droplevels()
drop_these_sports <- as.character(drop_these$Sport)
# select and filter
df <- oly12 %>%
as_tibble() %>%
select(Name, Country, Age, Height, Weight, Sex, Sport, Event) %>%
filter(!Sport %in% drop_these_sports) %>%
filter(Weight < 200) %>%
filter(complete.cases(.)) %>%
distinct() %>%
droplevels()Before leaving data carpentry, we should order the sports by some combination of height and weight. From our summaries earlier, we can see that the median height is 1.77 m and the median weight is 70 kg. The ratio of the two medians is about 1:40.
I’m going to order the factors by a weighted sum of the two quantities, w + 40h.
df <- df %>%
mutate(Sex = fct_reorder(Sex, Weight + 40*Height)) %>%
mutate(Sport = fct_reorder(Sport, Weight + 40*Height))A data carpentry file typically concludes by saving the data frame. I want to preserve the factor, so I’ll save to RDS format.
saveRDS(df, "data/0503-scatterplot-oly12-data.rds")Open the design script you initialized earlier.
A design file typically begins by reading the data prepared by the carpentry script.
library("tidyverse")
library("graphclassmate")
oly12 <- readRDS("data/0503-scatterplot-oly12-data.rds")
glimpse(oly12)
#> Observations: 8,993
#> Variables: 8
#> $ Name <fct> Lamusi A, A G Kruger, Jamale Aarrass, Maria Abakumova,...
#> $ Country <fct> People's Republic of China, United States of America, ...
#> $ Age <int> 23, 33, 30, 26, 27, 30, 23, 27, 28, 28, 22, 22, 25, 25...
#> $ Height <dbl> 1.70, 1.93, 1.87, 1.78, 1.82, 1.82, 1.87, 1.90, 1.73, ...
#> $ Weight <int> 60, 125, 76, 85, 80, 73, 75, 80, 60, 64, 62, 63, 69, 8...
#> $ Sex <fct> M, M, M, F, M, F, M, M, F, F, M, M, M, M, F, M, M, F, ...
#> $ Sport <fct> Judo, Athletics, Athletics, Athletics, Handball, Rowin...
#> $ Event <fct> "Men's -60kg", "Men's Hammer Throw", "Men's 1500m", "W...In geom_jitter() we add an alpha argument to add transparency to the
data markers.
p <- ggplot(data = oly12, aes(x = Height, y = Weight, col = Sex)) +
geom_jitter(alpha = 0.2) +
facet_wrap(vars(Sport), as.table = FALSE)
p
We need fewer columns to leave more room in the panel strips to be able to read the strip text (the sport). I’m also editing the breaks for tick marks to improve readability.
p <- p +
facet_wrap(vars(Sport), as.table = FALSE, ncol = 4) +
scale_x_continuous(breaks = seq(0, 3, by = 0.2)) +
scale_y_continuous(breaks = seq(0, 200, by = 50))
p
Next is the theme, labels, and manual color assignment using rcb()
from graphclassmate.
p <- p +
theme_graphclass() +
labs(x = "Height (m)",
y = "Weight (kg)",
title = "2012 Summer Olympics, Individual Athletes",
subtitle = "For sports with 50 or more athletes",
caption = "Source: VGAMdata package"
) +
scale_color_manual(values = c(rcb("mid_BG"), rcb("mid_Br")))
p
And write to file.
ggsave(filename = "0503-scatterplot-oly12.png",
path = "figures",
width = 8,
height = 16,
units = "in",
dpi = 300)The story: With the panels ordered by the overall “size” of the athletes (the weighted combination of height and weight), I find it interesting that water sports are at the two extremes. Water polo athletes have the largest median size and diving the smallest.
If we were to include this graph in a report, we would insert the following code chunk in the Rmd script.
```{r}
library("knitr")
include_graphics("../figures/0503-scatterplot-oly12.png")
```
1. gapminder
In the gapminder data, dollar values are in constant 2007 “international dollars”, a hypothetical unit of currency that has the same purchasing power parity (PPP) that the US dollar had in the US at the time. Constant dollars have been adjusted for inflation with respect to a particular year—in this case, 2007.
Scripts to initialize
explore/ 0503-scatterplot-gapminder-explore.R
carpentry/ 0503-scatterplot-gapminder-data.R
design/ 0503-scatterplot-gapminder.R
With the minimal instructions below, the graph you are trying to create is this one:
Data: gapminder from the gapminder package. If you want to learn more
about the data set, open its help page by running ? gapminder::gapminder
Explore script: write the code to determine the number of variables and for each variable, determine its
- type
- class
- levels (if categorical)
Carpentry script
- omit the continent column
- extract the data for the countries shown
- reorder the levels of the country factor by life expectancy
- recode country level
"Korea, Rep."to"South Korea" - drop unused levels
- write the data frame to
data/0503-scatterplot-gapminder-01.rds
Design script
- read the data frame you just saved
- create the the graph shown, with panels ordered as shown
- save it to file
figures/0503-scatterplot-gapminder-01.png
2. add two more variables
Continue the previous problem. In the design script,
- assign the color of the data marker to correspond year
- assign the size of the data marker to population
- save the figure to
figures/0503-scatterplot-gapminder-02.png
If you accomplish both, the graph should look like this,
3. one year, all countries
Carpentry script
- continue the same script
- start with the full gapminder data frame
- extract the data for the most recent year
- omit the year column
- omit data for the continent “Oceania”
- reorder the levels of the continent factor by life expectancy
- drop unused levels
- write the data frame to
data/0503-scatterplot-gapminder-03.rds
Design script
- read the data frame you just saved
- create the the graph shown, with panels ordered as shown
- save it to file
figures/0503-scatterplot-gapminder-03.png
Answer:
Doumont J-L (2009) Designing the graph. Trees, maps, and theorems: Effective communication for rational minds. Principiae, Kraainem, Belgium, 133–143 http://www.treesmapsandtheorems.com/
Wickham H and Grolemund G (2017) R for Data Science. O’Reilly Media, Inc., Sebastopol, CA https://r4ds.had.co.nz/