# Author Archives: Toni

## Gaps between bars

### Standard bar chart

A standard bar chart should have gaps between bars that are slightly narrower than the bars. The exceptions to this are the exception of histograms and clustered bar charts.

### Clustered bar chart

A clustered bar chart should have gaps between the clusters that are slightly wider than a single bar.

## Choosing the correct chart

There are eight common relationships that charts display. Prioritise what you want to highlight in the data and choose the chart type accordingly.

The eight common relationships within data are the following:

• comparisons of magnitude (size)
• time series
• ranking
• part-to-whole
• deviation
• distribution
• correlation
• spatial (maps will be covered separately in phase 4)

## Choice of data

Consider the message you want to communicate and choose your data accordingly. Your message might be better conveyed by deriving variables.

## Comparisons of magnitude (size)

To show:

• X is bigger than Y
• A is almost twice the size of B

Comparisons of size are shown most effectively as horizontal or vertical bars. Always begin the y-axis at zero.

### Small differences in magnitude, starting the y-axis at a non-zero value

If there are small differences between values sometimes it is necessary to start the y-axis at a non-zero value.

Always put a break in the y-axis if you do not start at zero.

#### Example of chart with a break in the y-axis

Use a dot (or other symbol) plot to make comparisons between values. The size of the visual element representing the data (dot position) is representative of the data value itself.

#### Example of chart using a dot plot to make comparisons

You can also show small differences between data by adjusting the deviation. This is changing what data can be seen from a chosen value (the deviation section has more information).

## Deviation

To show:

• number of times more than the average
• the difference from

Use a bar chart to plot deviation from a fixed value, or series of values.

#### Example of deviation where the value of data is most important

Gross disposable household income (GDHI) per head (£)
England, 2011

#### Example of deviation where the amount of change is most important

GDHI per head index comparison with England average
England, 2011

Use small multiples to plot deviation for multiple series. The axes should be identical for each small multiple.

## Distribution

To show:

• frequency
• distribution
• profile
• range
• concentration
• normal curve
• population pyramid
• shape

### For one variable

Use a histogram to show a distribution of data. Use small gaps between the bars to emphasise the profile of the data.

#### Example of histrogram chart showing one variable

Usually resident population aged 0 to 21 years
UK, 2013

### For two variables

Use a population pyramid to show the distribution of comparable data sets and highlight differences in the profile of the data.

### For more than two variables

To compare four variables population pyramids can be overlaid, with the least important data set displayed using an outline pyramid instead of bars.

#### Example of population pyramid showing multiple variables

Small multiple charts can also be used for multiple distributions. Use the same scale to enable comparison across charts.

#### Example of small multiples chart

Box-plots can also be used to compare distributions with two or more variables.

## Correlation

Correlation charts are often associated with causality and they should be used with caution.

Correlation can show:

• increases with
• relates to
• changes with
• varies with
• caused by

### Anscombe’s Quartet

Anscombe’s quartet is a powerful illustration of the drawback of relying solely on basic descriptive statistics to summarise data. The data in all four of the graphs in the quartet are virtually identical when using standard descriptive methods. Looking at your data before analysing it is something that Anscombe was passionate about.

#### Example of Anscombe’s Quartet

We are constantly improving based on research and best practice. Any significant changes to our guidance are available on the Updates page.

## Chart titles

Label charts as a figure and number them in order. Figures should have a main title and a statistical subtitle. Titles and subtitles should be concise and in sentence case.

### Main title

The main title should be descriptive, and tell the trend of the data or highlight the main story. Try to limit the number of words to no more than 10. This should make the description easier to read and avoid the text wrapping onto several lines, especially on mobile devices.

If you need to add context or detail to the chart, use annotations or support with your analysis.

### Statistical subtitle

The statistical subtitle should be as short as possible and must include the:

• statistical measure
• geographic coverage
• time period

You do not need to include these elements in the subtitle if they are already in the main heading.

### Writing chart titles to support your analysis

When writing your chart title and analysis:

• use chart titles to complement or build on, but not repeat section headings
• add further context and explanation of the chart’s message in your main text
• do not try and summarise everything the chart says in the title, but prioritise the main message

Take care not to use language in a title that you would not use in your analysis. Exaggerated language such as “greatest rise ever” may be more eye-catching, but use sparingly as it may appear sensationalist or could potentially be misinterpreted.

It can be useful to draw attention to a record level being recorded in the most recent data, but if a new record continues to be set every month, using the same title will lose its impact. Use sparingly and find another message to concentrate on instead.

#### Examples of how to write chart titles and subtitles

Your title can refer to a shorter period than shown on the chart. You can highlight an important short-term trend and give broader context by using a longer timeframe in your chart and analysis.

### If your chart has more than one message

If a chart has more than one narrative, choose the one that will be most relevant to users for the main title. Use annotations to draw attention to secondary messages, but do not try and explain every nuance in the chart when your analysis can provide more detail.

### Titles for other visual elements

Other types of visual content can communicate information. If you are using a flow chart or a map, the same titling principles apply. Use a descriptive title to tell the user what the story behind the image is, and use a statistical subtitle if appropriate.

#### Example of how to write a descriptive title for visual content

Sometimes a graphic may genuinely be one you wish your user to explore – there is no immediate story or message on display. For example, some of the interactive graphics coming from the Data Visualisation team may be in this space. In these rare cases, it is acceptable to use a title that encourages the reader to explore the graphic.

## Data order in tables

Group the data into meaningful subsets and make it clear in what order it should be read. Hierarchy and grouping can be shown by using white space and indenting column headings.

A table is made up of classification variables and data values, either of these can be used to order your table depending on the context.

If the table is not ranked by data value and the classification variables have a natural order, like age or geography, keep this order in the table.

Additional guidance is available on the recommended standard presentation order of statistics.

Put the variables that are most likely to be compared in columns, with the units, tens or hundreds beneath one another.

#### Where data are most likely to be compared between ages

Time should run from left to right or top to bottom.

## Number rounding

Always use a consistent level of precision, but use the lowest level possible for the intended user.

For the “inquiring citizen”, that is, a broader, less statistical audience :

260,000

For the “information forager”, for example, a local politician making decisions about future council tax charges:

264,300

For the “expert analyst”, for example, in a statistical journal discussing the detailed methodology behind the estimation, or in a situation where reproducibility is important:

264,337