Review

• A data visualisation consists of data symbols, guides, and labels.

• We produce data symbols by mapping data values to the visual features of a shape.

• Which visual feature we choose depends on:

  • Whether we have quantitative or qualitative data.
  • The accuracy and capacity of the visual feature.
  • The question we are interested in answering.

What is Data Visualisation?

• We have so far looked at the components that make up a data visualisation - data symbols, guides, and labels - in isolation from one another.

• In this topic, we consider the overall arrangement and layout of a data visualisation.

• Design ideas are aimed at improving aesthetic appeal.

• Design ideas also help to produce more consistent, coherent, and comprehensible data visualisations.

Graphic Design

• CRAP design

• {ggplot2} themes

• Fancier Text

• Clutter

CRAP Design

• Robin Williams popularised four standard design principles: contrast, repetition, alignment, and proximity.

  

Contrast

• The principle of contrast states that different elements of an image should either be obviously similar or obviously different.

  

Contrast

• Contrast should help the viewer to focus on the important elements of the data visualisation.

  

• Contrast is related to pre-attentive pop out.

Contrast

• An example of contrast is the use of colourful lines against the dull grey background.

• This helps the viewer to focus on the main data elements of the plot.

  

Repetition

• The principle of repetition states that related elements of an image should have similar visual features.

  

Repetition

• Repetition provides a structure and organisation that helps the viewer to navigate and comprehend an image.

  

• Repetition relates to Gestalt grouping.

Repetition

• An example of repetition is the use of consistent colours between the lines in the main plot and the lines in the legend.

  

Alignment

• The principle of alignment states that every element of an image should be aligned in some way with another element of the image.

  

Alignment

• Alignment helps to make an image look cleaner and more orderly, which again helps with navigation and reduces the cognitive load.

  

Alignment

• Examples of alignment are the left-justification of the plot title relative to the grey background and the legend title relative to the legend background.

  

Proximity

• The principle of proximity states that related elements of an image should be placed close together.

  

• Proximity relates to Gestalt grouping.

Proximity

• Proximity also helps with structure and organisation, particularly to help the viewer to identify elements that belong together as a group.

  

Proximity

• An example of proximity is the placement of the elements of the plot legend

• The elements of the legend are close to each other, with an empty margin between them and the main plot.

  

CRAP Design

• We can apply each of the CRAP principles of graphic design to both critically assess an existing data visualisation and to suggest ideas for improvement.

  

Contrast

• We can increase the contrast between elements:

  • The title and the lines and labels that represent the data.
  • The background, grid lines, and axis labels.

• There is now more emphasis on the title and the data.

  

Alignment

• We can align multiple elements with both the left and right ends of the data lines.

• There is now a simpler and cleaner structure.

  

Repetition

• We can repeat the colours in lines and labels.

• This creates clear groupings of these elements.

  

Proximity

• We can increase the proximity of the labels and the lines.

• This removes the need for a separate legend, further simplifying the plot.

  

Themes in {ggplot2}

• Producing a complete data visualisation involves making a lot of design decisions.

• Part of the convenience of {ggplot2} comes from the fact that it makes a lot of decisions for us.

• If we want to control the contrast, repetition, alignment, and proximity of visual elements, we need to take control of these decisions.

• Themes provide a way to modify the default {ggplot2} decisions.

• {ggplot2} makes a lot of decisions about scales, colours, and labels.

  ggplot(crimeGender) +
      geom_line(aes(yearDate, count, colour=gender)) +
      labs(title="Number of Incidents") 

  

Themes in {ggplot2}

• The theme() function allows individual details to be adjusted.

• Functions like theme_bw() and theme_dark() change a whole host of defaults at once.

• Functions like theme_bw() and theme_dark() change a whole host of defaults at once.

  ggplot(crimeGender) +
      geom_line(aes(yearDate, count, colour=gender)) +
      labs(title="Number of Incidents") +
      theme_bw()

  

The theme() function

• The arguments to theme() are different theme settings, e.g., plot.title to control the style of the plot title.

• The value supplied for each setting depends on the type of setting, e.g., plot.title controls a text element so we provide a value using element_text().

• Functions like element_text() have arguments to control text settings, like family amd size.

• Functions like element_text() have arguments to control text settings, like family and size.

  ggplot(crimeGender) +
      geom_line(aes(yearDate, count, colour=gender)) +
      labs(title="Number of Incidents") +
      theme(plot.title=element_text(size=20, face="bold"))

  

• element_rect() provides settings for borders and backgrounds, like panel.background.

  ggplot(crimeGender) +
      geom_line(aes(yearDate, count, colour=gender)) +
      labs(title="Number of Incidents") +
      theme(panel.background=element_rect(colour="black", fill=NA))

  

• element_line() provides settings for lines, like panel.grid.

  ggplot(crimeGender) +
      geom_line(aes(yearDate, count, colour=gender)) +
      labs(title="Number of Incidents") +
      theme(panel.grid=element_line(colour="black", linewidth=1))

  

• element_blank() is used to remove a plot element entirely.

  ggplot(crimeGender) +
      geom_line(aes(yearDate, count, colour=gender)) +
      labs(title="Number of Incidents") +
      theme(panel.background=element_blank())

  

The theme() function

• Some themes settings are more general and provide defaults for more specific settings.

  • For example, there are axis.title, axis.title.x, and axis.title.x.bottom settings.

• Some theme settings are neither text, nor lines, nor rectangles.

  • For example, legend.position is a character vector that specifies the location of the plot legend: "left", "right", "bottom", "top", or "none".

Example

• The code for the modified line plot controls many theme settings.

  

Fancier Text in {ggplot2}

• Standard text labels in R can only have a single font and style.

• The {ggtext} package provides support for changing the style within a label.

  

Two things are required:

• basic markup (markdown or HTML/CSS) in the text.

• element_markdown() within theme().

  ggplot(crimeGender) +
      geom_line(aes(yearDate, count, colour=gender)) +
      labs(title=paste0('<span style="color: ', cols[2], 
                        '">**Males**</span>',
                        " Responsible for More Offences than ",
                        '<span style="color: ', cols[1], 
                        '">**Females**</span>'),
           caption=paste0('Total number of offences by youth ',
                          'aged between 14 and 16 from 2010 ',
                          'to 2020 in New Zealand.<br>',
                          '**Source:** New Zealand Ministry ',
                          'of Justice *"Youth Justice Indicator ',
                          'Report"*, 2021.')) +
      theme(plot.title=element_markdown(),
            plot.caption=element_markdown(hjust=0),
            legend.position="none")

Effective Data Visualisation

• We have emphasised the importance of mapping data values to visual features that facilitate mapping back to the data values.

Clutter

• Visual elements that have no connection to the data create clutter.

• Clutter may distract attention from the data symbols.

• Clutter may add to the complexity of the data visualisation (the number of elements to fixate on).

  

Clutter

• Not all embellishments are necessarily evil.

  

An effective data visualisation will have minimal clutter.

• Tufte calls clutter chart junk and suggests maximising the data-ink ratio.

• Many other data visualisation guidelines advocate minimising clutter.

• There are some dissenting voices.

  • There is some evidence that embellishments may improve engagement and recall.

  • There is also some evidence that making the viewer work harder may improve accuracy and retention.

Summary

• Simple applications of contrast, repetition, alignment, and proximity can have a significant impact on the overall effectiveness of a data visualisation.

  • Reduce cognitive load.
  • Reduce clutter.
  • Direct attention.

• {ggplot2} themes allow control over many of the details of a data visualisation.

• The {ggtext} package allows finer control over text labels.

• Less clutter is generally better.

Exercise

• Can you identify any problems in the design of this visualisation?

• Can you suggest any ways to improve it?

  Figure 1 from “Entropy Ordered Shapes as Bivariate Glyphs”, Halliman et al (2023).

Exercise

• Can you identify any weaknesses in the data visualisation below in terms of contrast, repetition, alignment, and proximity?

  ggplot(crimeLevel) +
      geom_line(aes(year, count, colour=level))

  

Exercise

• Can you create this modified version of the data visualisation?

  HINT: These are mostly just theme() settings.

  

Exercise

• Can you see anything wrong with this data visualisation?

  New Zealand Listener December 9-15 2023.