1.

10 Exercise: Install and start using R

Using R for Data to Insight
This is a late breaking initiative to make analyses in R available in this run of Data to Insight.
The intended audience is people who already have had some experience with installing
software and some experience of coding, or at least of using some command- driven
system.
R is made up of a base system and many thousands of additional packages which give it
enormous capability. Almost all of our analyses will use functions from Tom Elliot’s R
package iNZightPlots which in turn draw on many other packages.
Warning: Coding is entirely unforgiving. If you get anything wrong, however small (e.g.
missing a bracket, misspelling a name, using a lower-case letter when the name has an
upper-case letter or vice versa as R is case sensitive), you will just get error messages. So be
very careful, and even then expect to make mistakes.
1. Install R: Go to https://cran.r-project.org/mirrors.html and click on a CRAN mirror
site near you. Download and install R (download versions available for Windows, Mac
and Linux).
Warning: If you are a Mac user and may also want to use iNZight just use the
version of R installed while installing iNZight. Installing a second version is likely to
create problems for you. The iNZight installation will already have installed the
packages in number 3. below.

2. Start up R.

3. Install the R packages we will be using in the course by copying the following 2 lines
of code and pasting them into the R Console window (not in R Studio).
install.packages(c('iNZightPlots', 'FutureLearnData'), dependencies = TRUE,
repos = c('http://r.docker.stat.auckland.ac.nz/R', 'https://cran.rstudio.com'))
If it asks you, “Would you like to create a personal library to install packages into? “,
say, “Yes”.

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 1 of 3

 4. When that has completed, paste the following 2 lines into the R Console window
library(iNZightPlots)
library(FutureLearnData)
You will get error messages if these packages have not installed.

5. Now try the following: (Paste lines of code, or even several lines of code at a time,
into the R Console window. See what they do.
# R CODE COMMENTARY

These first 2 lines have to be run every time you start up

R and want to use the functionality in iNZightPlots, or the
data in FutureLearnData.

library(iNZightPlots) Load the iNZightPlots package

library(FutureLearnData) Load the FutureLearnData package (contains all the data
sets for the course)

data(package="FutureLearnData") Tell me about all the data sets in FutureLearnData

data(nhanes_1000) Make the data set nhanes_1000 in FutureLearnData

available for analysis

nhanes_1000[1:10, 1:8] Show me the first 10 rows and 8 columns of nhanes_1000

head(nhanes_1000) Show me the top rows of nhanes_1000

tail(nhanes_1000) Show me the bottom rows of nhanes_1000

names(nhanes_1000) Give me the names of all of the variables in nhanes_1000

iNZightPlot(Race3, data=nhanes_1000) Plot the variable named Race3 in nhanes_1000

iNZightPlot(Height, data=nhanes_1000) Plot the variable named Height in nhanes_1000

getPlotSummary(Height, data=nhanes_1000) Get me a Summary of the variable named Height in
nhanes_1000

4. Ask for plots and summaries of other variables whose names you can see in the
names list.

5. When you have finished, close R. When it asks “Save Workspace image? “, click,
“No”.

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 2 of 3

To discuss issues related to this Exercise,
go to https://gitter.im/iNZightVIT/d2i-R-discussion
To be able to post to the list you will have to set up a (free) account on Github
https://github.com/login
If your question relates to an Exercise, say which one you are talking about!

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 3 of 3

1.15 Exercise: Import data into R
(R version of Exercise 1.15)

From Exercise 1.10 (R version) you have already seen how to make data sets in the
FutureLearnData package available for analysis but we will reiterate the general pattern soon.
The # character in R: If you type or paste a line into the R Console window, R will ignore everything
that comes after a “#” character. So # tells R that what follows is a comment left for human
readers, not an instruction for R itself.
We will use this in the following as we talk about the pattern for making the data in a package, in
our case the FutureLearnData package available for analysis.

library(FutureLearnData) # Load the package FutureLearnData

data(package= "FutureLearnData") # give me info about the data in the package FutureLearnData
# I can copy and paste from this to get the names of data sets exactly right
data(olympics100m) # data(dataset name) makes it available for use
olympics100m # saying the name of something causes it to display
# OK to do here as this particular dataset is small
# Otherwise use commands from Exercise 1.10 for displaying small parts of the data set
Olympics100m # this name is wrong because of the capital "O" so will give an error
data(package= “FutureLearnData”) # curly quotation marks “ from Word, not straight ones, ", so error
# This whole block of lines can be copied and pasted as code. Try it

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 1 of 3

Reading csv and tab-separated text files into R
It is simple to read rectangular data sets in csv or tab-separated text file formats into R. We will do
it now.
1. Download the file Census at School-500.csv from
https://www.stat.auckland.ac.nz/~wild/d2i/FutureLearn/

2. Download the file olympics100m.txt from

https://www.stat.auckland.ac.nz/~wild/d2i/FutureLearn_TabTxt/

3. Now try the following: (Paste lines of code, or even several lines of code at a time, into the
R Console window. See what they do.

# R CODE COMMENTARY

# Import the file Census at School-500.csv read.csv is asking R to read a csv file
file.choose() is telling R to throw up a
cas_500 = read.csv(file.choose(), header = TRUE) browser window that will allow you to
navigate to wherever you have stored
Census at School-500.csv and open the file
header = TRUE tells R that this file has a
header line containing the names of the
variables
cas_500 = tells R to store the result as
cas_500

cas_500[1:5, 1:9] Show me the first 5 rows and 9 columns of

cas_500

names(cas_500) Give me the names of all of the variables in

cas_500
library(iNZightPlots) Need to load iNZightPlots package if not
already done this session
Plot the variable named armspan in
iNZightPlot(armspan, data= cas_500) cas_500

# Now import the file olympics100m.txt As above but to read the tab-separated text
file we use read.table, not read.csv. We
Olymp_imp = read.table(file.choose(), header = TRUE, sep="\t") include sep="\t" to tell R to look for tab
characters as the separators between data
fields
We store the result as store it as
Olymp_imp

names(Olymp_imp) Give me the names of all of the variables in

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 2 of 3
 Olymp_imp
iNZightPlot(YEAR, TIME, data= Olymp_imp) Plot YEAR, TIME in Olymp_imp (gives a
scatter plot of y=TIME versus x=YEAR)

?read.table Show me the help file for the function

read.table
?read.csv
Show me the help file for the function
read.csv.
In this case the same help file covers both
of these closely related functions

[Note: Most actions in R are invoked by calling an R function. Function calls in R are of the
form:
function.name(list of function parameters separated by commas)
When you look at help files you will note in the “Usage” paragraph that a function will often
have a large number of parameters. You do not need to include any parameters in your call to a
function if that parameter is set equal to a value in this paragraph. That assigned value is the
default value. You do not need to include any parameter that has a default in your call unless
you want to change its value from the default to something else.]

4. Try some variations of the above, e.g. plotting new variables, reading another data file.

5. When you have finished, close R. When it asks “Save Workspace image?”, click, “No”.

To discuss issues related to this Exercise,

go to https://gitter.im/iNZightVIT/d2i-R-discussion
To be able to post to the list you will have to set up a (free) account on Github
https://github.com/login
If your question relates to an Exercise, say which one you are talking about!

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 3 of 3

2.5 Exercise: Categorical variables – R version
# R code Output
# Setup
library(iNZightPlots)
library(FutureLearnData)
data(nhanes_1000)
# Plot the variable Race3
# Because Race3 is categorical we get a bar chart

iNZightPlot(Race3, data=nhanes_1000)

# Get a summary for a variable (Race3)

# Equiv. of Get Summary in iNZight

getPlotSummary(Race3, data=nhanes_1000)

# Equivalent of Get Inference in iNZight

getPlotSummary(Race3, data=nhanes_1000,
summary.type="inference", inference.type="conf")

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 1 of 4

# Colour by a variable (Race3) (default colour palette)

iNZightPlot(Race3, data=nhanes_1000, colby=Race3)

COMMENTARY
# Create a new variable Race3.reord to re-order Race3 R calls a categorical variable a “factor”
# with the categories in frequency order
Show me the levels of Race3 (I can also see in
levels(nhanes_1000$Race3) the graph). Output is …
[1] "Asian" "Black" "Hispanic" "Mexican" "Other"
"White"
I can see what the frequency order should be
from the graph. (This can be done generally with
nhanes_1000$Race3.reord = code but the code is too complex to do at this stage)
factor(nhanes_1000$Race3, levels = c("White", So I’ll make Race3.reord from Race3 and put
"Black", "Mexican", "Hispanic", "Asian", "Other") ) them in the order I want. (Getting the number of
levels and spelling exactly right is crucial)

iNZightPlot(Race3.reord, data=nhanes_1000)

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 2 of 4

# We’ll do this again putting the levels of Education into Distribution of Education

a sensible order

30
25
iNZightPlot(Education, data=nhanes_1000)

Proportion (%)

20
15
10
5
0
ad
e e rad ol e
Gr rad eG cho lleg
thG hS Co
8th 11 lleg Hig me
9_ Co So

Education
285 missing values

levels(nhanes_1000$Education) [1] "8thGrade" "9_11thGrade" "CollegeGrad"

"HighSchool" "SomeCollege"

# Create a new variable to re-order Education

nhanes_1000$Education.reord =
factor(nhanes_1000$Education, levels =
c("8thGrade","9_11thGrade","HighSchool","SomeCo
llege","CollegeGrad") )
[1] "8thGrade" "9_11thGrade" "HighSchool"
levels(nhanes_1000$Education.reord) "SomeCollege" "CollegeGrad"

Distribution of Education.reord

iNZightPlot(Education.reord, data=nhanes_1000)
30
25
Proportion (%)

20
15
10
5
0

ad
e e ol e
rad
Gr rad cho o lleg eG
8th thG hS eC lleg
11 Hig m Co
9_ So

Education.reord
285 missing values

iNZightPlot(Education.reord, COMMENTARY
data=nhanes_1000,colby=Education.reord) Colour by Education.reord
# Now change the colour palette to rainbow colours
Col.fun has to be a colour palette function
iNZightPlot(Education.reord, There are lots of colour palette functions in R,
data=nhanes_1000,colby=Education.reord, many you have to install other packages to get.
col.fun=rainbow) rainbow() is a generally available colour palette

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 3 of 4

library(colorspace)
iNZightPlot(Education.reord, Using the rainbow_hcl colour function from the
data=nhanes_1000,colby=Education.reord, colorspace package
col.fun=rainbow_hcl)

 Try repeating the above using other choices for variables and settings

If you want to try installing some other R packages, in the R menus

Go Packages > Install packages . You will probably be asked to choose a CRAN mirror site.

Then you will be shown a list of packages to choose from.

Installing the package viridis and then loading it [via library(viridis)] will give you access to
the colour functions: viridis, magma, and inferno

To discuss issues related to this Exercise,

go to https://gitter.im/iNZightVIT/d2i-R-discussion
To be able to post to the list you will have to set up a (free) account on Github
https://github.com/login
If your question relates to an Exercise, say which one you are talking about!

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 4 of 4

2.10 Exercise: Numeric variables – R version
#R Code Commentary
Because Pulse is numeric we get a dot plot
# Plot a numeric variable (Pulse)

iNZightPlot(Pulse, data=nhanes_1000)

# Changing the size of the dots

iNZightPlot(Pulse, data=nhanes_1000, cex.dotpt=.4)

iNZightPlot(Pulse, data=nhanes_1000, cex.dotpt=2)

# Get a Summary for Pulse

getPlotSummary(Pulse, data=nhanes_1000)

# Equivalent of Get Inference for Pulse

getPlotSummary(Pulse, data=nhanes_1000,
summary.type="inference",
inference.type="conf")

Coloured by Age
iNZightPlot(Height, data=nhanes_1000)

# Colour points by Age

iNZightPlot(Height, data=nhanes_1000, colby=Age)

# Change colour palette to rainbow

iNZightPlot(Height, data=nhanes_1000, colby=Age,

col.fun=rainbow)

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 1 of 2

iNZightPlot(Height, data=nhanes_1000)

# Change Plot type to histogram

iNZightPlot(Height, data=nhanes_1000, plottype=

"hist")

# Control the number of bins (class intervals)

iNZightPlot(Height, data=nhanes_1000, plottype=

"hist", hist.bins=10)

# Get a list of all the other things that can be changed

in plots

inzpar() # This list is complete

?inzpar # This documentation is not complete

 Try doing more things like the above but using other variables and settings

To discuss issues related to this Exercise,

go to https://gitter.im/iNZightVIT/d2i-R-discussion
To be able to post to the list you will have to set up a (free) account on Github
https://github.com/login
If your question relates to an Exercise, say which one you are talking about!

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 2 of 2

2.13 Exercise: Comparing groups – R version
We are going to use the gapminder_2008 data in the FutureLearnData package and look at
the values of ChildrenPerWoman for each of the countries broken out by Region.

# R CODE COMMENTARY or OUTPUT

# Setup Commentary
library(iNZightPlots)
library(FutureLearnData)
Make gapminder_2008 inside
data(gapminder_2008) FutureLearnData available for analysis

names(gapminder_2008) Useful for checking on the spellings of

variable names

iNZightPlot(ChildrenPerWoman, data=gapminder_2008) Dot plot for ChildrenPerWoman

# Now break out by Region

iNZightPlot(ChildrenPerWoman, Region,
data=gapminder_2008)

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 1 of 2

# Get Summary of ChildrenPerWoman broken out
by Region

getPlotSummary(ChildrenPerWoman, Region,
data=gapminder_2008)

# Colour by Region

iNZightPlot(ChildrenPerWoman, Region,
data=gapminder_2008, colby=Region)

# Try also
iNZightPlot(ChildrenPerWoman, Region, data=gapminder_2008,
colby=Region, cex.text=.3)
iNZightPlot(ChildrenPerWoman, Region, data=gapminder_2008,
colby=Region, hide.legend = TRUE)

# Colour by Infantmortality

iNZightPlot(ChildrenPerWoman, Region,
data=gapminder_2008, colby=Infantmortality)

 What do you see in the last graph?

 Also try colouring by other variables you think might help explain the Regional
differences.

To discuss issues related to this Exercise,

go to https://gitter.im/iNZightVIT/d2i-R-discussion
To be able to post to the list you will have to set up a (free) account on Github
https://github.com/login
If your question relates to an Exercise, say which one you are talking about!
Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 2 of 2
2.15 Exercise: Time travel – R version
In this Exercise we are going to use the gapminder data set that contains data for all years
whereas gapminder_2008 (that we used in the last Exercise) only had the data for 2008.
We will …
1. Start by looking at the plot of ChildrenPerWoman by Region
o This is not sensible as we are getting lots of points for the same country
sometimes going as far back as 1952
2. Make a separate graph like this for every year (as a tiled set of plots)
3. See how to look at just one of these plots
4. See how to “play through the years” by displaying the graph for each year in turn

#R Code Commentary OR Output

# Setup Commentary

library(iNZightPlots)
library(FutureLearnData)
data(gapminder) Use gapminder NOT gapminder_2008
names(gapminder)

iNZightPlot(ChildrenPerWoman, Region, data=gapminder) Plot ChildrenPerWoman by Region

# Plot ChildrenPerWoman by Region subset by Year_cat

iNZightPlot(ChildrenPerWoman, Region, g1=Year_cat, data=gapminder)

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 1 of 2

 Remind ourselves how the levels
levels(gapminder$Year_cat) of Year_cat are represented
# Display just the plot for the year 2000 [1] "[1952]" "[1956]" "[1960]" …

iNZightPlot(ChildrenPerWoman, Region, g1=Year_cat, Choose the level value

g1.level="[2000]", data=gapminder) corresponding to the year 2000

# Do it again for 2004

iNZightPlot(ChildrenPerWoman, Region, g1=Year_cat,

g1.level="[2004]", data=gapminder)

# Now put it in a loop and do it for every year, i.e. for every level of Year_cat

for (k in levels(gapminder$Year_cat)) iNZightPlot(ChildrenPerWoman,Region, g1=Year_cat,

g1.level=k, data=gapminder)

# Do not display a new plot UNTIL you have clicked on the on plot window

old.value = devAskNewPage(TRUE) # save current plotting behaviour and ask for new behaviour

for (k in levels(gapminder$Year_cat))
iNZightPlot(ChildrenPerWoman,Region, g1=Year_cat, g1.level=k, data=gapminder)

devAskNewPage(old.value) # Reset the plotting behaviour back to the way it was before

Commentary
# Play the plots, but with a 2 second delay between plots
This time there are 2 lines of
for (k in levels(gapminder$Year_cat)) { code to be run at each step so we
iNZightPlot(ChildrenPerWoman,Region, g1=Year_cat, g1.level=k, have to put them in “{ .. }”
data=gapminder) brackets so that both lines get
Sys.sleep(2) run
}

To discuss issues related to this Exercise,

go to https://gitter.im/iNZightVIT/d2i-R-discussion
To be able to post to the list you will have to set up a (free) account on Github
https://github.com/login
If your question relates to an Exercise, say which one you are talking about!

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 2 of 2

3.5 Exercise: Relationships between categorical
variables – R version
This exercise will enable you to construct graphs of two categorical variables as discussed in
the previous video. The skills addressed are:

1. Creating plots of two categorical variables (when the predictor variable has only 2
groups).
2. Making a side-by-side bar charts and separate bar charts for two categorical
variables.
3. Filtering out unwanted groups

We will use the nhanes_1000 dataset from the FutureLearnData package.

#R Code Output and/or Commentary

# Setup
library(iNZightPlots)
library(FutureLearnData)
data(nhanes_1000)

Education with levels in a sensible

# Recreate Education.reord order
nhanes_1000$Education.reord = factor(nhanes_1000$Education, (This was done in Exercise 2.5)
levels =
c("8thGrade","9_11thGrade","HighSchool","SomeCollege","CollegeGrad"))

iNZightPlot(Education.reord, data=nhanes_1000)

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 1 of 3

# Education.reord by Gender (side by side barcharts)

iNZightPlot(Education.reord, Gender, data=nhanes_1000)

# Education subset by Gender (also coloured by Gender)

iNZightPlot(Education.reord, g1=Gender, data=nhanes_1000,

colby=Gender)

# Display just the female plot

iNZightPlot(Education.reord, g1=Gender, g1.level="female",

data=nhanes_1000, colby=Gender)

# Display just the male plot

(Separate plots)

iNZightPlot(Education.reord, g1=Gender, g1.level="male",

data=nhanes_1000, colby=Gender)

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 2 of 3

# Plot Education.reord subset by AgeDecade

iNZightPlot(Education.reord, g1=AgeDecade, data=nhanes_1000,

colby=AgeDecade)

# Filter out unhelpful data (Under 20s)

# Make a subset of the data without “under 20s” & call it Temp

# (This is an example of “filtering” the data) (Warning: there is a leading space on all
of AgeDecade’s level names– a data bug)
Temp = subset(nhanes_1000, AgeDecade!=" 0-9" & AgeDecade
!=" 10-19") # “!=” is read as “is not equal to”
Output
table(Temp$AgeDecade) 0-9 10-19 20-29 30-39 40-49 50-59 60-69 70+
0 0 146 110 144 111 05 66
It still has the empty levels
# This will remove the empty levels
Temp$AgeDecade=factor(Temp$AgeDecade)

# Replot the data using the subset of the data called Temp
iNZightPlot(Education.reord, g1=AgeDecade, data=Temp,
colby=AgeDecade)

 Now construct similar graphics using other categorical variables whose behaviour
you may be interested in

To discuss issues related to this Exercise,

go to https://gitter.im/iNZightVIT/d2i-R-discussion
To be able to post to the list you will have to set up a (free) account on Github
https://github.com/login
If your question relates to an Exercise, say which one you are talking about!

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 3 of 3

3.9 Exercise: Relationships between numeric
variables – R version
This exercise will enable you to construct graphs of two numeric variables as discussed in
the previous video. The skills addressed are:

1. Creating a scatterplot of two numeric variables (when the predictor variable has
more than groups).
2. Adding a trend line (if suitable) and label any interesting points.
3. Train your eyes to observe and draw an envelope around the scatter.

We will use the gapminder_2008 dataset in the FutureLearnData package.

Creating a scatterplot of two numeric variables

When we call iNZightPlot(x,y) using two numeric variables x and y we will get a scatterplot.
The 1st variable (x) gets plotted against the horizontal axis and the 2nd variable (y) is plotted
against the vertical axis. So we should put the predictor variable first and the outcome
variable second.

[Note: I’ve just seen that have reversed the order of the variables in this exercise from the
way it was in the (older) iNZight version. The order below is more natural. The change
affects nothing that matters in the exercise but the graphs are “the other way around” from
the iNZight version.]

#R Code Commentary

# Setup
library(iNZightPlots)
library(FutureLearnData)

data(gapminder_2008)

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 1 of 4

# Plot y=EnergyUsePerPerson by x=CO2emissions

iNZightPlot(CO2Emissions, EnergyUsePerPerson,
data=gapminder_2008)

# Note the reversal in the order of the names between the call and
ordinary English usage for what we want to do!
Recall: the plot call is in the order
iNZightPlot(x,y,…)

# Label the 4 most extreme points by the variable Country

iNZightPlot(CO2Emissions, EnergyUsePerPerson,
data=gapminder_2008, locate.extreme=4, locate=Country)

# Add a trend line

iNZightPlot(CO2Emissions, EnergyUsePerPerson,
data=gapminder_2008, trend="linear")

# Changing from the default Axes labels and title

iNZightPlot(CO2Emissions, EnergyUsePerPerson,
data=gapminder_2008, xlab="CO2 emissions per person ",
ylab="Energy use per person", main="Energy use versus CO2
emissions")

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 2 of 4

# Put on a trend line but coloured red, instead of the default (blue)

iNZightPlot(CO2Emissions, EnergyUsePerPerson,
data=gapminder_2008, trend="linear",
col.trend=list(linear="red"))

Commentary
# Label specified data points

ids = (1:nrow(gapminder_2008))[gapminder_2008$Country %in% Tell me the row numbers of all the

rows of gapminder_2008 for which
c("Qatar","Iceland","Sweden")]
the value of Country is in the set listed
in c(…).
Store these row numbers in ids

“locate.id” is used to tell iNZightPlot

iNZightPlot(CO2Emissions, EnergyUsePerPerson, the row numbers of the points it
data=gapminder_2008, locate.id=ids, locate=Country) should label
“locate=Country” tells iNZightPlot to
use the contents of the variable
Country to label those points

The options added to the plots above can all be used together. Try putting several of them
together in the same call to iNZightPlot. There are lots of others you will find out about over
time, e.g. if you are putting a line on your plot then adding “, lwd=2” to your call (without
the “”) will double the thickness of the line.
Try to identify more or other countries (spelling and lower/upper case is critical).
Try working with other variables.

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 3 of 4

Observations of scatterplots
Not all scatterplots show obvious trends with uniform scatter.
For each scatterplot below, train your eye to look for trends, clusters, outliers and
draw an envelope around the scatter. You may want to print them out so that you can
draw on the above graphs.
For each graph, what do you notice? Post one of your findings on the discussion.

To discuss R issues related to this Exercise,

go to https://gitter.im/iNZightVIT/d2i-R-discussion
If your question relates to an Exercise, say which one you are talking about!
Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 4 of 4
4.7 Exercise: Techniques for scatterplots
– R version
This exercise will enable you to become more proficient in creating scatterplots with
iNZightPlot. You will learn how to apply the most suitable trend line and use
techniques to overcome perceptual problems.

The skills addressed are:

1. Create a scatterplot of two numeric variables and apply a suitable trend line.
2. Use techniques such as jittering, transparency and running quartiles to deal
with overprinting.

We are going to explore the relationship between variables Age and Weight of
people in the nhanes_1000 dataset in the FutureLearnData package.

# R code Output and/or Commentary

# Setup
library(iNZightPlots)
library(FutureLearnData)
data(nhanes_1000)

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 1 of 4

# Plot Weight vs Age

iNZightPlot(Age, Weight,data=nhanes_1000)

# Add a trend line

iNZightPlot(Age, Weight, data=nhanes_1000, trend="linear")

# Get some summary information (result depends on trends fitted)

getPlotSummary(Age, Weight, data=nhanes_1000, trend="linear")

getPlotSummary(Age, Weight, data=nhanes_1000, trend="linear",

summary.type="inference")

The inference results make no

sense here as the real trend is
clearly not linear

# Add more trend curves and a smoother

iNZightPlot(Age,Weight,data=nhanes_1000, trend=c("linear",
"quadratic", "cubic") , smooth=.25)

The value given for the smooth

(between 0 and 1) controls the
level of smoothing. Bigger
numbers correspond to more
smoothing

# Make all the lines thicker, all solid lines, and change line colours

iNZightPlot(Age,Weight,data=nhanes_1000, trend=c("linear",
"quadratic", "cubic") , smooth=.25, lwd=2,
lty.trend=list(linear=1,quadratic=1,cubic=1),
col.trend=list(linear="red", quadratic="yellow", cubic="blue"),
col.smooth="green")

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 2 of 4

# Scatter plot coloured by Height and with Transparency

iNZightPlot(Age, Weight, data=nhanes_1000, colby=Height, alpha=.3)

# alpha (0 to 1) controls transparency. Smaller for more transparent

# Make the points smaller

iNZightPlot(Age, Weight, data=nhanes_1000, colby=Height, alpha=.3,

cex.pt=.2)

# Create a median smoother in black

iNZightPlot(Age, Weight, data=nhanes_1000, colby=Height, alpha=.3,

quant.smooth=.5, col.smooth="black")

# Create a median and quartile smoothers all in black

iNZightPlot(Age, Weight, data=nhanes_1000, colby=Height, alpha=.3,

quant.smooth=c(.25,.5,.75), col.smooth="black")

iNZightPlot(Age, HomeRooms, data=nhanes_1000)

# Jitter in the vertical (y) direction

iNZightPlot(Age, HomeRooms, data=nhanes_1000, jitter="y")

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 3 of 4

  Play some more with these settings and try other variables
 For even more settings, type ?inzpar into R to get help on the inzpar, or type inzpar
to just get a complete list (last time I looked the help file wasn’t entirely complete)

To discuss issues related to this Exercise,

go to https://gitter.im/iNZightVIT/d2i-R-discussion
To be able to post to the list you will have to set up a (free) account on Github
https://github.com/login
If your question relates to an Exercise, say which one you are talking about!

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 4 of 4

4.11 Exercise: Advanced scatterplots for deeper analysis – R
version
This exercise will enable you to explore more complicated relationships between variables and the explore
the effects of a third and fourth variable, enabling you to view changes over time.

The skills addressed are:

1. Create a scatterplot of two numeric variables, subset by a 3rd variable.

2. Explore the effect of a third and fourth variable using colour and size.

We will use the gapminder dataset (but not gapminder_2008).

Create a scatterplot of two numeric variables, subset by a 3rd variable

We are going to explore the relationship between the variables Infantmortality and ChildrenPerWoman of
countries in the Gapminder dataset over time.

#R Code Output and/or Commentary

# Setup
library(iNZightPlots)
library(FutureLearnData)
data(gapminder)
# Scatterplot of Infantmortaility against ChildrenPerWoman

iNZightPlot(ChildrenPerWoman,Infantmortality ,
data=gapminder)

# Subset by Year_cat

iNZightPlot(ChildrenPerWoman,Infantmortality, g1=Year_cat,
data=gapminder)

# Change size and colour of points

iNZightPlot(ChildrenPerWoman,Infantmortality,g1=Year_cat,
data=gapminder, colby=Region, sizeby=Populationtotal)

# Show results for 1972 only

iNZightPlot(ChildrenPerWoman,Infantmortality,g1=Year_cat,
g1.level="[1972]",data=gapminder, colby=Region,
sizeby=Populationtotal)

# Darker background (often easier to see some of the lighter dots)

iNZightPlot(ChildrenPerWoman,Infantmortality, g1=Year_cat,
g1.level="[1972]",data=gapminder, colby=Region,
sizeby=Populationtotal, bg="darkgray")

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 2 of 4

# Try transparency and smaller points (removed garkgray)

iNZightPlot(ChildrenPerWoman,Infantmortality, g1=Year_cat,
g1.level="[1972]",data=gapminder, colby=Region,
sizeby=Populationtotal, alpha=.45, cex.dotpt=.5)

# Try subsetting by different years, e.g. g1.level="[1976]",

# Label some of the extreme points (ask for 4)

iNZightPlot(ChildrenPerWoman,Infantmortality, g1=Year_cat,
g1.level="[1972]",data=gapminder, colby=Region,
sizeby=Populationtotal, alpha=.45, cex.dotpt=.5,
locate.extreme=4, locate=Country)

# Label some specific countries

ids = (1:nrow(gapminder))[gapminder$Country %in% c("United
States of America","China","Brazil", "India")]
iNZightPlot(ChildrenPerWoman,Infantmortality, g1=Year_cat,
g1.level="[1972]",data=gapminder, colby=Region,
sizeby=Populationtotal, alpha=.45, cex.dotpt=.5,
locate.id=ids, locate=Country)

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 3 of 4

# Allow a little more room on left to accommodate label

iNZightPlot(ChildrenPerWoman,Infantmortality, g1=Year_cat,
g1.level="[1972]",data=gapminder, colby=Region,
sizeby=Populationtotal, alpha=.45, cex.dotpt=.5,
locate.id=ids, locate=Country, xlim=c(0,9))

# Subset by a fourth variable (Region)

iNZightPlot(ChildrenPerWoman, Infantmortality, g1=Year_cat,

g2=Region, g2.level="America", data=gapminder, colby=Region)

# Play through the years

for (k in levels(gapminder$Year_cat)) {
iNZightPlot(ChildrenPerWoman,Infantmortality, g1=Year_cat,
g1.level=k, data=gapminder, colby=Region,
sizeby=Populationtotal, alpha=.45, cex.dotpt=.5,
locate.id=ids, locate=Country)
Sys.sleep(1)
}

 Play some more with these settings and try other variables
 For even more settings, type ?inzpar into R to get help on the inzpar, or type inzpar
to just get a complete list (last time I looked the help file wasn’t entirely complete)

To discuss issues related to this Exercise,

go to https://gitter.im/iNZightVIT/d2i-R-discussion
To be able to post to the list you will have to set up a (free) account on Github
https://github.com/login
If your question relates to an Exercise, say which one you are talking about!

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 4 of 4

6.12 Exercise: Inference with iNZight – R version
This exercise will enable you to make comparisons between sub-groups allowing for
sampling error.

Background understanding: see Step 6.9.

The skills addressed in this Exercise are:

 Use iNZightPlot to get inferential mark-ups of plots so that you can make visual
comparisons between sub-groups allowing for sampling error.
 To obtain numerical confidence limits for true between-group differences.

We will use the nhanes_1000 dataset from the FutureLearnDatasets package.

#R code Output and/or Commentary

# Setup
library(iNZightPlots)
library(FutureLearnData)
data(nhanes_1000)
names(nhanes_1000)

# Plot BMI by AgeDecade

iNZightPlot(AgeDecade, BMI, data=nhanes_1000)

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 1 of 3

# Add inference information

iNZightPlot(AgeDecade, BMI, data=nhanes_1000,

inference.type=c("comp","conf"),
inference.par="mean")

# Commentary (on this side to save space)

# You can squash your plot window vertically so that it is

easier to see how much overlap there is between each
age group.

# What do you see here? The thick black lines are called
‘comparison intervals’ and are the lines that we
look at when are the lines that we look at when
observing any overlap. The thin red lines are the
individual confidence intervals for each
mean/median.

# View detailed inferential information (Normal Theory)

getPlotSummary(AgeDecade, BMI, data=nhanes_1000,

summary.type="inference", inference.type="conf")

etc

We’ll now do this with a pair of

Categorical variables
First need to do some filtering to get rid of
# Filter out under 20s decades with no General Health data
Temp=subset(nhanes_1000,AgeDecade !=" 0-9" &
AgeDecade !=" 10-19")
Temp$AgeDecade=factor(Temp$AgeDecade)

# Reorder the HealthGen variable and create HealthGen.reord with the levels in
Temp$HealthGen.reord = factor(Temp$HealthGen, levels a sensible order
= c("Excellent","Vgood","Good","Fair","Poor") )

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 2 of 3

# Plot as separate bar charts with inference information

iNZightPlot(HealthGen.reord, g1=AgeDecade, data=Temp,

inference.type=c("comp","conf") )

# Plot as side by side without inference comparison

iNZightPlot(HealthGen.reord, AgeDecade, data=Temp,

inference.type=c("comp","conf") )

 Play some more with these settings and try other variables

To discuss issues related to this Exercise,

go to https://gitter.im/iNZightVIT/d2i-R-discussion
To be able to post to the list you will have to set up a (free) account on Github
https://github.com/login
If your question relates to an Exercise, say which one you are talking about!

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 3 of 3

Exercise: Time Series for a single variable
– R version
In this exercise we will use iNZightTS package to create a Time Series plot and get
the Additive and Multiplicative Decomposition for it.

We will use the Week8_AverageVisitorsQuarterly dataset from the FutureLearnData

package.

The skills addressed are to use the iNZightTS package to:

1. Generate a Time Series plot and a Seasonal plot for a single numeric variable.
2. Get an Additive and Multiplicative Decomposition.
3. Make a forecast.

Generate a Time Series plot and a Seasonplot for a single numeric variable

The data we are using shows us the number of visitors from different countries who
are currently staying in New Zealand. We will investigate the changes in the number
of Australian visitors over time.

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 1 of 4

8.6 Time Series for a single variable

# R Code Output and/or Commentary

Commentary
# Install the iNZightTS package
This installation only has to be done once
install.packages(c('iNZightTS'), dependencies =
TRUE, repos =
c('http://r.docker.stat.auckland.ac.nz/R',
'https://cran.rstudio.com'))

# Load the Time Series Library

Load packages in every new R session in which you
library(iNZightTS)
want to use them
library(FutureLearnData)

# List the course datasets and select dataset

data(package="FutureLearnData") Look at summary of the data sets to ensure you
have the name exactly correct

data(week8_AverageVisitorsQuarterly) Load the data set (which is in the FutureLearnData

package)
head( week8_AverageVisitorsQuarterly)
List the top few rows

Commentary
Australia = iNZightTS
(week8_AverageVisitorsQuarterly, var=2) Create Time Series object for the Australian series
Australia is the 2nd variable in the dataset
# Plot the data -- t controls smoothing

rawplot( Australia , t=25) t controls the smoothing (t must be betw. 0 and 100)

# Now experiment with different values of t

# If you are on Windows and using R alone (not

RStudio) try …

rawplot( Australia , t=25,animate=TRUE)

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 2 of 4

# Seasonal plot – additive

seasonplot( Australia )

# Seasonal plot – multiplicative

seasonplot( Australia , multiplicative=TRUE)

# Decomposition plot

decompositionplot( Australia, t=20)

# Recomposed plot

decomp=decompositionplot( Australia, t=20)

recompose(decomp, animate=FALSE)

# If you are on Windows and using R alone (not

RStudio) try …
recompose(decomp, animate=TRUE)

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 3 of 4

# Calculate a forecast

forecastplot( Australia )

Commentary
# Let’s establish this pattern for another country

China = iNZightTS (week8_AverageVisitorsQuarterly, var=3) Create Time Series object for the China
series. China is the 3nd variable in the
dataset week8_AverageVisitorsQuarterly
rawplot( China , t=20)
decompositionplot( China, t=20) Now you can start plotting …

# etc …

 Repeat what we have done above for any other country that interests you and try
to interpret the patterns you see as has been done in the video
o Skim-read the iNZight version for the commentary that is missing here. (This
document just concentrates on how the code works)
[In the next Exercise we will start comparing series from different countries.]

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 4 of 4

Exercise: Time Series analysis for more than one series
– R version
In the previous exercise we only looked at the visitors from one country alone. Now we
want to see the graphs for more than one, to be able to compare their visitor numbers.

This exercise will enable you to use iNZight to compare several time series by viewing them
simultaneously in two different ways.

We will again use the Week8_AverageVisitorsQuarterly dataset from the FutureLearnData

package.

# R Code Output and/or Commentary

# Set up
library(iNZightTS)
library(FutureLearnData)

data(week8_AverageVisitorsQuarterly)

# See the variables in this dataset

head(week8_AverageVisitorsQuarterly)

# See how these work …

We are going to use this idea to specify the
c(2,5,9)
column numbers corresponding to the
c(2,4:6, 8)
countries we want to look at

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 1 of 3

Aus_USA = Create Time Series object for the set of countries
iNZightTS(week8_AverageVisitorsQuarterly, we want to look at
var=c(2,9)) selecting columns 2 and 9 (which correspond to
Australia and USA).
Let’s call it Aus_USA

# Separate plots for Aus_USA

multiseries(Aus_USA, t=20)

Create Time Series object for the whole set of

countries we want to look at
ALL = iNZightTS(week8_AverageVisitorsQuarterly,
 selecting all columns from 2 to 9.
var=c(2:9))  Let’s call it ALL

# Separate plots for ALL

multiseries(ALL, t=20)

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 2 of 3

# Separate plots for Aus_USA

compareplot(Aus_USA, t=30)

# Separate plots for ALL

compareplot(ALL, t=30)

 Repeat what we have done above for any other combinations of countries that
interest you and try to interpret the patterns you see as has been done in the video
o Skim-read the iNZight version for the commentary that is missing here. (This
document just concentrates on how the code works)
 and for exploration questions

Data to Insight: An Introduction to Data Analysis The University of Auckland | Page 3 of 3