DataViz: Ensemble Graphics
Heather Kitada Smalley
Ensemble Graphics
These examples modified from Chapter 12 Ensemble Graphics and Case Studies in Graphical Data Analysis with R,by Antony Unwin
Motivating Example 1: Coffee
#install.packages("pgmm")
library(pgmm)
data("coffee")
head(coffee)## Variety Country Water Bean Weight Extract Yield ph Value Free Acid
## 1 1 mexico 8.939999 156.6 33.5 5.80 32.7
## 2 1 mexico 7.400000 157.3 32.1 5.81 30.8
## 3 1 guatemal 9.740000 152.9 33.1 5.26 36.7
## 4 1 honduras 10.400000 174.0 31.5 5.61 34.2
## 5 1 salvador 10.540000 145.1 35.2 5.77 31.8
## 6 1 salvador 10.000000 156.4 34.5 5.83 32.6
## Mineral Content Fat Caffine Trigonelline Chlorogenic Acid
## 1 3.80 15.2 1.13 1.03 5.38
## 2 3.71 15.0 1.25 1.01 5.13
## 3 4.15 16.1 1.21 1.05 5.94
## 4 3.94 15.8 1.06 0.94 5.87
## 5 4.09 15.2 1.11 0.99 5.09
## 6 3.88 15.4 1.20 0.81 5.30
## Neochlorogenic Acid Isochlorogenic Acid
## 1 0.40 0.79
## 2 0.32 0.97
## 3 0.24 0.76
## 4 0.39 0.59
## 5 0.49 0.72
## 6 0.43 0.69names(coffee)## [1] "Variety" "Country" "Water"
## [4] "Bean Weight" "Extract Yield" "ph Value"
## [7] "Free Acid" "Mineral Content" "Fat"
## [10] "Caffine" "Trigonelline" "Chlorogenic Acid"
## [13] "Neochlorogenic Acid" "Isochlorogenic Acid"1.1 Exploration:
Is that any missing data?
A unicorn!
#install.packages("naniar")
library(naniar)## Warning: package 'naniar' was built under R version 3.6.2vis_miss(coffee)
1.1.1 Visualize One Categorical Variable: Bar Charts
What is the distribution of coffee varieties?
library(tidyverse)
# barcharts for categorical vars
ggplot(coffee, aes(factor(Variety)))+
geom_bar()
# more arabia than robustaWhat is the distribution of coffee origin countries?
length(unique(coffee$Country))## [1] 29# 29 different countries (with odd spellings)
ggplot(coffee, aes(Country))+
geom_bar()+
coord_flip()
1.1.2 Visualize Multiple Categorical Variables: Mosaic plots
What is the distribution of coffee varieties across country?
#mosaic for categorica
#install.packages("vcd")
library(vcd)## Warning: package 'vcd' was built under R version 3.6.2## Loading required package: gridmosaic(Variety~Country, data=coffee)
# only one country has both varietiesHere’s another example of a mosaic plot that is built into this package:
#example
mosaic(HairEyeColor, shade=TRUE, legend=TRUE)
1.1.3 Visualize Continuous Variables: Standardized boxplots
Look for outliers
# distribution of continuous variables
boxplot(scale(coffee[,3:14]))
# standardizes1.1.4 Visualize Relationships between Two Continuous Variables: Pairs plots
# scatterplot matrix of continuous
pairs(coffee[,3:14], pch=16)
It can be hard to see a pairs plot with too many variables. Let’s try fewer.
pairs(coffee[,3:7], pch=16)
I also really like the GGally package for pairs plots.
#install.packages("GGally")
library(GGally)## Warning: package 'GGally' was built under R version 3.6.2## Registered S3 method overwritten by 'GGally':
## method from
## +.gg ggplot2ggpairs(coffee[,3:7])
1.2 Polishing/Explanation
- Labels
- Names of Varieties
- Names of Variables
- Consistent colors
1.2.1 Names of Varieties
coffee<-within(coffee, Type <- ifelse(Variety==1,
"Arabica", "Robusta"))
head(coffee)## Variety Country Water Bean Weight Extract Yield ph Value Free Acid
## 1 1 mexico 8.939999 156.6 33.5 5.80 32.7
## 2 1 mexico 7.400000 157.3 32.1 5.81 30.8
## 3 1 guatemal 9.740000 152.9 33.1 5.26 36.7
## 4 1 honduras 10.400000 174.0 31.5 5.61 34.2
## 5 1 salvador 10.540000 145.1 35.2 5.77 31.8
## 6 1 salvador 10.000000 156.4 34.5 5.83 32.6
## Mineral Content Fat Caffine Trigonelline Chlorogenic Acid
## 1 3.80 15.2 1.13 1.03 5.38
## 2 3.71 15.0 1.25 1.01 5.13
## 3 4.15 16.1 1.21 1.05 5.94
## 4 3.94 15.8 1.06 0.94 5.87
## 5 4.09 15.2 1.11 0.99 5.09
## 6 3.88 15.4 1.20 0.81 5.30
## Neochlorogenic Acid Isochlorogenic Acid Type
## 1 0.40 0.79 Arabica
## 2 0.32 0.97 Arabica
## 3 0.24 0.76 Arabica
## 4 0.39 0.59 Arabica
## 5 0.49 0.72 Arabica
## 6 0.43 0.69 Arabica1.2.2 Names of Variables
names(coffee)<-abbreviate(names(coffee), 8)
names(coffee)## [1] "Variety" "Country" "Water" "BeanWght" "ExtrctYl" "ph Value"
## [7] "FreeAcid" "MnrlCntn" "Fat" "Caffine" "Trignlln" "ChlrgncA"
## [13] "NchlrgnA" "IschlrgA" "Type"1.3 Graphics
1.3.1 Bargraph
a<- ggplot(coffee, aes(x=Type)) +
geom_bar(aes(fill=Type))+
scale_fill_manual(values=c("grey70", "red"))+
guides(fill=FALSE)+
ylab("")## Warning: `guides(<scale> = FALSE)` is deprecated. Please use `guides(<scale> =
## "none")` instead.a
1.3.2 Scatterplot
b<-ggplot(coffee, aes(x=Fat, y=Caffine, colour=Type))+
geom_point(size=3)+
scale_colour_manual(values=c("grey70", "red"))
b
1.3.3 Parallel Coordinate Plot
#install.packages("GGally")
library(GGally)
c<-ggparcoord(coffee[order(coffee$Type),], columns=3:14,
groupColumn="Type", scale="uniminmax",
mapping=aes(size=1))+
xlab("") + ylab("")+
theme(legend.position="none")+
scale_colour_manual(values=c("grey", "red"))+
theme(axis.ticks.y=element_blank(),
axis.text.y=element_blank())
c
1.4 Ensemble Graphic
#install.packages("gridExtra")
library(gridExtra)##
## Attaching package: 'gridExtra'## The following object is masked from 'package:dplyr':
##
## combinegrid.arrange(arrangeGrob(a, b, ncol=2, widths=c(1, 2)),
c, nrow=2)
Case Study: Fertility
Mothers in this dataset had two or more children.
### Fertility dataset
#install.packages("AER")
library(AER)
data("Fertility")
head(Fertility)## morekids gender1 gender2 age afam hispanic other work
## 1 no male female 27 no no no 0
## 2 no female male 30 no no no 30
## 3 no male female 27 no no no 0
## 4 no male female 35 yes no no 0
## 5 no female female 30 no no no 22
## 6 no male female 26 no no no 402.1 Exploration
2.1.1 What is the distribution of age?
“As expected, the numbers of women with two or more children rise with age, although it is perhaps surprising that the pattern is so regular before dropping off for the two oldest cohorts”
ggplot(Fertility, aes(x=age))+
geom_bar()
2.1.2 Is the distribution of age different across demographic groups?
The same pattern can be seen across all race groups.
ggplot(Fertility, aes(x=age))+
geom_bar(binwidth=1)+
facet_wrap(~afam+hispanic+other, scales="free_y")## Warning: Ignoring unknown parameters: binwidth
2.1.3 Does whether or not the mother worked affect number of children?
In these data, a majority of the women either didn’t work or worked, with some in-between.
We can look at the proportions:
with(Fertility, prop.table(table(work)))## work
## 0 1 2 3 4 5
## 0.4717813190 0.0056036819 0.0066796516 0.0053366529 0.0084428283 0.0040800459
## 6 7 8 9 10 11
## 0.0078420131 0.0022854540 0.0116707375 0.0025446292 0.0086352463 0.0014254636
## 12 13 14 15 16 17
## 0.0195363120 0.0029844416 0.0046847880 0.0047515452 0.0152874096 0.0025485561
## 18 19 20 21 22 23
## 0.0043863438 0.0010170663 0.0186370526 0.0014568787 0.0031297368 0.0011702153
## 24 25 26 27 28 29
## 0.0092046463 0.0064911606 0.0112034368 0.0010170663 0.0059492488 0.0007539642
## 30 31 32 33 34 35
## 0.0128762949 0.0007461104 0.0088512256 0.0010367008 0.0022147698 0.0045041507
## 36 37 38 39 40 41
## 0.0180912140 0.0019673753 0.0046141038 0.0025210678 0.0289176687 0.0011741422
## 42 43 44 45 46 47
## 0.0054976556 0.0014686594 0.0057332695 0.0055408515 0.0047044225 0.0023011616
## 48 49 50 51 52
## 0.0140700715 0.0047122763 0.0261256450 0.0063694268 0.18542414412.2 Graphics
Since we are using the same dataset and similar geometries we can use the same “base” and then add modifiers to it.
2.2.1 Age and Weeks Worked
p0<-ggplot(Fertility)+geom_bar(binwidth=1)+ylab("")## Warning: Ignoring unknown parameters: binwidthp1<- p0+aes(x=age)
p2<-p0+aes(x=work)+xlab("Weeks worked in 1979")
p1
p2
2.2.2 Other Variables
k<-ggplot(Fertility)+geom_bar()+ylab("")+ylim(0, 250000)
p3<-k+aes(x=morekids) + xlab("has more children")
p4<-k+aes(x=gender1) + xlab("first child")
p5<-k+aes(x=gender2) + xlab("second child")
p6<-k+aes(x=afam) + xlab("African American")
p7<-k+aes(x=hispanic) + xlab("Hispanic")
p8<-k+aes(x=other) + xlab("other race")2.2.3 Now combine!
grid.arrange(arrangeGrob(p1, p2, ncol=2, widths = c(3,3)),
arrangeGrob(p3, p4, p5, p6, p7, p8, ncol=6),
nrow=2, heights=c(1.25, 1))
2.3 Spinogram plots
Note that these are different than colored bar charts because the widths also show the proportions
2.3.1 More than two kids by age
doubledecker(morekids~age, data=Fertility,
gp=gpar(fill=c("grey90", "forestgreen")),
spacing=spacing_equal(0))
2.3.2 More than two kids by genders of previous children
doubledecker(morekids~gender1+gender2, data=Fertility,
gp=gpar(fill=c("grey90", "forestgreen")))
2.3.3 More than two kids by age and genders of previous children
doubledecker(morekids~age+gender1+gender2, data=Fertility,
gp=gpar(fill=c("grey90", "forestgreen")),
spacing=spacing_dimequal(c(0.1, 0, 0, 0)))
3. Case Studies
3.1 Moral statistics of France
#install.packages("HistData")
library(HistData)## Warning: package 'HistData' was built under R version 3.6.2data(Guerry)
head(Guerry)## dept Region Department Crime_pers Crime_prop Literacy Donations Infants
## 1 1 E Ain 28870 15890 37 5098 33120
## 2 2 N Aisne 26226 5521 51 8901 14572
## 3 3 C Allier 26747 7925 13 10973 17044
## 4 4 E Basses-Alpes 12935 7289 46 2733 23018
## 5 5 E Hautes-Alpes 17488 8174 69 6962 23076
## 6 7 S Ardeche 9474 10263 27 3188 42117
## Suicides MainCity Wealth Commerce Clergy Crime_parents Infanticide
## 1 35039 2:Med 73 58 11 71 60
## 2 12831 2:Med 22 10 82 4 82
## 3 114121 2:Med 61 66 68 46 42
## 4 14238 1:Sm 76 49 5 70 12
## 5 16171 1:Sm 83 65 10 22 23
## 6 52547 1:Sm 84 1 28 76 47
## Donation_clergy Lottery Desertion Instruction Prostitutes Distance Area
## 1 69 41 55 46 13 218.372 5762
## 2 36 38 82 24 327 65.945 7369
## 3 76 66 16 85 34 161.927 7340
## 4 37 80 32 29 2 351.399 6925
## 5 64 79 35 7 1 320.280 5549
## 6 67 70 19 62 1 279.413 5529
## Pop1831
## 1 346.03
## 2 513.00
## 3 298.26
## 4 155.90
## 5 129.10
## 6 340.733.2 Airbags and car accidents
#install.packages("DAAG")
library(DAAG)## Loading required package: lattice## Warning: package 'lattice' was built under R version 3.6.2##
## Attaching package: 'DAAG'## The following object is masked from 'package:survival':
##
## lung## The following object is masked from 'package:car':
##
## vifdata(nassCDS)
head(nassCDS)## dvcat weight dead airbag seatbelt frontal sex ageOFocc yearacc yearVeh
## 1 25-39 25.069 alive none belted 1 f 26 1997 1990
## 2 10-24 25.069 alive airbag belted 1 f 72 1997 1995
## 3 10-24 32.379 alive none none 1 f 69 1997 1988
## 4 25-39 495.444 alive airbag belted 1 f 53 1997 1995
## 5 25-39 25.069 alive none belted 1 f 32 1997 1988
## 6 40-54 25.069 alive none belted 1 f 22 1997 1985
## abcat occRole deploy injSeverity caseid
## 1 unavail driver 0 3 2:3:1
## 2 deploy driver 1 1 2:3:2
## 3 unavail driver 0 4 2:5:1
## 4 deploy driver 1 1 2:10:1
## 5 unavail driver 0 3 2:11:1
## 6 unavail driver 0 3 2:11:23.3 Athletes’ blood measurements
library(DAAG)
data(ais)
head(ais)## rcc wcc hc hg ferr bmi ssf pcBfat lbm ht wt sex sport
## 1 3.96 7.5 37.5 12.3 60 20.56 109.1 19.75 63.32 195.9 78.9 f B_Ball
## 2 4.41 8.3 38.2 12.7 68 20.67 102.8 21.30 58.55 189.7 74.4 f B_Ball
## 3 4.14 5.0 36.4 11.6 21 21.86 104.6 19.88 55.36 177.8 69.1 f B_Ball
## 4 4.11 5.3 37.3 12.6 69 21.88 126.4 23.66 57.18 185.0 74.9 f B_Ball
## 5 4.45 6.8 41.5 14.0 29 18.96 80.3 17.64 53.20 184.6 64.6 f B_Ball
## 6 4.10 4.4 37.4 12.5 42 21.04 75.2 15.58 53.77 174.0 63.7 f B_Ball3.4 Marijuana arrests
#install.packages("effects")
library(effects)## Warning: package 'effects' was built under R version 3.6.2## Use the command
## lattice::trellis.par.set(effectsTheme())
## to customize lattice options for effects plots.
## See ?effectsTheme for details.data("Arrests")
head(Arrests)## released colour year age sex employed citizen checks
## 1 Yes White 2002 21 Male Yes Yes 3
## 2 No Black 1999 17 Male Yes Yes 3
## 3 Yes White 2000 24 Male Yes Yes 3
## 4 No Black 2000 46 Male Yes Yes 1
## 5 Yes Black 1999 27 Female Yes Yes 1
## 6 Yes Black 1998 16 Female Yes Yes 03.5 Crohn’s disease
#install.packages("robustbase")
library(robustbase)## Warning: package 'robustbase' was built under R version 3.6.2##
## Attaching package: 'robustbase'## The following object is masked from 'package:DAAG':
##
## milk## The following object is masked from 'package:survival':
##
## heartdata("CrohnD")
head(CrohnD)## ID nrAdvE BMI height country sex age weight treat
## 1 19908 4 25.22 163 c1 F 47 67 placebo
## 2 19909 4 23.80 164 c1 F 53 64 d1
## 3 19910 1 23.05 164 c1 F 68 62 placebo
## 4 20908 1 25.71 165 c1 F 48 70 d2
## 5 20909 2 25.95 170 c1 F 67 75 placebo
## 6 20910 2 28.70 168 c1 F 54 81 d13.6 Footballers in the four major European leagues
#install.packages("SportsAnalytics")
library(SportsAnalytics)
data(EURO4PlayerSkillsSep11)
head(EURO4PlayerSkillsSep11)## Name League Team Number Position Positions
## 1 Tobias Sippel Bundesliga 1. FC Kaiserslautern 1 Goalkeeper GK!
## 2 Kevin Trapp Bundesliga 1. FC Kaiserslautern 29 Goalkeeper GK!
## 3 Jan Simunek Bundesliga 1. FC Kaiserslautern 2 Defender CB!
## 4 Leon Jessen Bundesliga 1. FC Kaiserslautern 3 Defender SB!, WB
## 5 Martin Amedick Bundesliga 1. FC Kaiserslautern 5 Defender CB!
## 6 Matthias Abel Bundesliga 1. FC Kaiserslautern 6 Defender CB!, SB
## Birthday Nationality Age Height Weight InjuryTolerance Foot Side Attack
## 1 <NA> German NA 180 79 B R B 30
## 2 (08/01/1990) German 21 189 83 B R B 30
## 3 <NA> Czech NA 189 79 B R B 56
## 4 (11/06/1986) Danish 25 182 75 B L L 70
## 5 <NA> German NA 194 85 B R B 60
## 6 (22/06/1981) German 30 188 82 B R R 63
## Defence Balance Stamina TopSpeed Acceleration Response Agility
## 1 81 83 69 70 72 84 80
## 2 82 84 69 70 72 84 83
## 3 78 82 84 78 76 78 78
## 4 66 79 83 82 80 76 75
## 5 78 87 83 74 75 79 73
## 6 77 84 82 76 74 78 72
## DribbleAccuracy DribbleSpeed ShortPassAccuracy ShortPassSpeed
## 1 52 58 56 65
## 2 54 58 56 65
## 3 71 76 71 70
## 4 75 78 72 74
## 5 71 73 73 73
## 6 71 72 70 73
## LongPassAccuracy LongPassSpeed ShotAccuracy ShotPower ShotTechnique
## 1 63 69 44 83 44
## 2 63 68 44 83 44
## 3 68 70 66 80 64
## 4 76 79 64 81 65
## 5 71 71 66 81 65
## 6 71 73 64 80 65
## FreeKickAccuracy Curling Header Jump Technique Aggression Mentality
## 1 44 48 44 83 54 64 78
## 2 44 48 44 84 53 64 82
## 3 65 66 82 80 72 67 75
## 4 70 76 72 73 76 78 76
## 5 63 67 82 80 71 66 81
## 6 63 68 80 76 71 65 77
## KeeperSkills Teamwork ConditionFitness WeakFootAccuracy WeakFootFrequency
## 1 81 76 6 4 4
## 2 82 79 6 4 4
## 3 50 75 6 4 4
## 4 50 78 6 5 5
## 5 50 80 7 5 5
## 6 50 77 6 4 43.7 Decathlon
#install.packages("GDAdata")
library(GDAdata)
data("Decathlon")
head(Decathlon)## Totalpoints DecathleteName Nationality m100 Longjump Shotput Highjump
## 1 8559 Torsten Voss DDR 10.66 8.00 14.73 2.06
## 2 8504 Uwe Freimuth DDR 11.10 7.66 16.30 1.94
## 3 8440 Siegfried Wentz BRD 11.21 7.22 15.84 2.09
## 4 8409 Aleksandr Nevski SU 10.95 7.35 14.99 2.08
## 5 8381 John Sayre USA 10.86 7.41 14.22 2.00
## 6 8366 Vadim Podmaryov SU 11.09 7.56 15.28 2.08
## m400 m110hurdles Discus Polevault Javelin m1500 yearEvent P100m Plj Psp
## 1 48.28 14.50 43.28 4.9 61.28 268.80 1985 938 1061 773
## 2 48.46 14.77 47.72 4.9 68.26 270.56 1985 839 975 870
## 3 47.75 14.28 45.52 4.7 69.66 278.38 1985 814 866 841
## 4 49.29 14.76 46.12 4.6 68.16 261.09 1985 872 898 789
## 5 49.98 14.84 46.08 5.3 67.68 277.07 1985 892 913 742
## 6 50.00 14.89 48.58 4.6 67.46 272.31 1985 841 950 807
## Phj P400m P110h Ppv Pdt Pjt P1500
## 1 859 896 911 880 732 757 752
## 2 749 887 878 880 823 863 741
## 3 887 921 939 819 778 884 691
## 4 878 848 879 790 790 861 804
## 5 803 816 869 1004 789 854 699
## 6 878 815 863 790 841 851 730