Minor adjustments

• Sometimes we would like to add a title to a plot or change the labels.
  • This is achieved by labs
• Sometimes we may wish to change the general look
  • This is achieved by the theme function and functions in the ggthemes package.

Change labels

economics%>%
  ggplot(aes(x=psavert,y=uempmed))+
  geom_point()

Change labels

economics%>%
  ggplot(aes(x=psavert,y=uempmed))+
  geom_point()+labs(title = 'Savings v Duration')+
  xlab('Savings Rate')+
  ylab('Medium Duration of Unemployment')

Change theme

economics%>%
  ggplot(aes(x=psavert,y=uempmed))+
  geom_point()+theme_classic()

Change theme

economics%>%
  ggplot(aes(x=psavert,y=uempmed))+
  geom_point()+theme_bw()

One from ggthemes

economics%>%
  ggplot(aes(x=psavert,y=uempmed))+
  geom_point()+theme_economist()

Another from ggthemes

economics%>%
  ggplot(aes(x=psavert,y=uempmed))+
  geom_point()+theme_wsj()

Theme

• You can customise your own themes using the theme function.
• There are many guides for using this
• One can be found here.

Annotation

• Remember that plots tell a story.
• Sometimes it helps to add text to a plot to help tell this story.
• This can be done with the annotate function.

Annotate

economics%>%
  ggplot(aes(x=psavert,y=uempmed))+
  geom_point()+ annotate('text',
   x=4,y=24,label='Bad Times!',size=5,col='red')

Many time series on one plot

• It is common to see multiple time series on a single plot.
• This can be achieved using the group aesthetic.
• We will do this with the txhousing data.
• Suppose we are looking at sales and listings since 2010 for Houston only.

Houston Data

txhousing%>%
  filter(city=='Houston',date>2010)%>%
  select(date,sales,listings)->houston_sales_listings
houston_sales_listings%>%
  pivot_longer(cols = -date, 
               names_to = 'variable', 
               values_to = 'value')%>%
  ggplot(aes(x=date,y=value,group=variable))+
  geom_line()

Houston Data

Beware

• Plotting time series measured in different units on the same plot is very very risky.
• Something commonly seen is a plot with two time series but with different y axes.
• For some debate on this issue see this discussion .
• Crossing points can be manipulated by arbitrarily changing the scale.

Crossing points

Aspect Ratio

• Another issue when looking at time series plots is the aspect ratio.
• The aspect ratio is the ratio of the width to the height of the plot.
• For time series plot larger aspect ratios can make trends look smaller.
• Aspect ratio can be controlled through the coord_fixed function

Aspect Ratio

houston_sales_listings%>%
  ggplot(aes(x=date,y=sales))+
  geom_line()+
  coord_fixed(ratio=0.002)
houston_sales_listings%>%
  ggplot(aes(x=date,y=sales))+
  geom_line()+coord_fixed(0.0001)

Aspect Ratio

Banking to 45

• An old rule of thumb suggested by Cleveland is banking to 45 degrees
• Find the slopes of every line joining a point at time $t$ to the the point at time $t+1$.
• Set the aspect ratio so that the median of these is 45 degrees.
• It is only a rough guide and has recently been called into question.

Alternatives

• If we are looking for the relationship between two variables it is often better to look at a scatterplot.
• A problem with this is that the dimension of time is lost.
• Time can however be represented using color and the geom_path geometry.

Scatter plot

houston_sales_listings%>%
  ggplot(aes(x=listings,y=sales))+
  geom_point()

Path plot

houston_sales_listings%>%
  ggplot(aes(x=listings,y=sales, col=date))+
  geom_path()+scale_color_viridis_c()

Modelling

• Often visualisation is conducted with modelling in mind
• Scatterplots can be visualised with a model fit as well
• This is done using geom_smooth
• We can illustrate using the mpg data.

Scatterplot

With fitted curve

ggplot(mpg,aes(x=displ,y=cty))+
  geom_point()+geom_smooth()

Smooth fit

• The fitted line by default comes from the LOcal Estimated Scatterplot Smoothing (LOESS) method.
• This technique combines the idea of nearest neighbours with regression.
• Nearest neighbours are found (along the x-axis)
• A constant, linear or quadratic regression is fit to the nearest neighbours.

Nearest Neighbours

Nearest Neighbours

Local fit

Nearest Neighbours

Details

• In the LOESS algorithm, the smoothing parameter $\alpha$ is defined as the proportion of observations used as nearest neigbours.
  • If $\alpha =0.2$ and $n=20$ then $k=4$ nearest neighbours are used.
  • By default $\alpha =0.75$ in R.
• It is common to use weighted regression whereby closer neighbours are given more influence.
• LOESS is not ideal for large datasets.

Linear fit

Linear regression can be used instead of LOESS

ggplot(mpg,aes(x=displ,y=cty))+
  geom_point()+geom_smooth(method = 'lm')

## `geom_smooth()` using formula 'y ~ x'

With Colour

ggplot(mpg,aes(x=displ,y=cty,col=drv))+
  geom_point()+geom_smooth(method = 'lm')

## `geom_smooth()` using formula 'y ~ x'

Confidence bands

• The grey ribbons give an indication of uncertainty around the estimated line.
• These relate to uncertainty around the estimate of the regression slope (or LOESS curve).
• Since regression includes a noise term, observations can easily lie outside the confidence ribbons.

Visualisation on the web

• Plots these days are often looked at on a website or at least on a computer.
• This is in contrast to the recent past when most plots were eventually printed onto paper.
• This allows the user to interact with plots.
• An example of this is the plotly software.

Plotly and ggplot

• There is an R package called plotly which allows plotly to be easily used with ggplot.
• Simply store the result of ggplot in a variable and then run the function ggplotly

houston_sales_listings%>%
  ggplot(aes(x=listings,y=sales, col=date))+
  geom_point()+scale_color_viridis_c()->g
ggplotly(g)

Plotly

More plotly

Code

houston_sales_listings%>%
  mutate(Era=ifelse(date>2014,
                    'Post 2014',
                    'Pre 2014'))%>%
  ggplot(aes(x=listings,y=sales, col=Era))+
  geom_point()->g
ggplotly(g)

Summary

• Other interesting tools include the gganimate package.
  • As the name suggests this allows for easy animation.
• Also the shiny package
  • This allows for the design of interactive web apps.
• You now have a strong base in R to teach yourself these how to use the tools.

Final Exercise

Create this plot

Solution

txhousing%>%
  filter(month==4,city%in%c('Houston','San Antonio',
                            'Dallas','Austin'))%>%
  ggplot(aes(x=sales,y=median,col=year,label=year))+
  geom_path()+
  geom_text(size=2,color='black')+
  scale_color_viridis_c()+
  labs(title = 'Median Price v Number of Sales in 
       April for four Texan Cities')+
  xlab('Sales (Number of Houses)')+
  ylab('Median House Price ($US)')+
  facet_wrap(~city,scales = 'free')