Communicating Numbers II - Charitable Charts

Tables vs. Graphs:

Tables are ideal to look up individual values. The ‘tabular’ format of rows and columns facilitates tracing the information.

Graphs however reveal the shape of the data, which can not easily be gleaned by looking at a table.

Quantitative vs. Categorical Data

Quantitative data is numerical information that can be measured or counted.

Qualitative data is descriptive information about characteristics that are difficult to describe numerically. These qualities may be represented by a name, symbol, or a number code.

Quantitative and qualitative data are often used together to get a full picture of a population. For example in a survey information about occupation (quality) and income (quantity) complement each other.

Qualitative data can still be categorized by its level of measurement.

• Nominal data → are categories without order, like colors or names. They are differentiated by their name (nominally) and analysis is limited to recording the types and frequency.

• Ordinal data → has a natural order or ranking (unsatisfied, neutral, satisfied). But the difference between these ranks is either not measurable, unequal or meaningless.

• Interval data → consists of quantitative data but collected into equal intervals like $[10°C-20°C], [20°C-30°C]$. Interval data can be ranked and the difference between data points is measurable. But intervals lack a true zero point, in the case of degrees $0°C$ does not mean the absence of temperature. Also, a multiplication or division is not meaningful, $20°C\,(68°F)$ is not twice as much as $10°C\,(50°F)$, because the ratio changes depending on the scale.

Data Display

Numbers become meaningful when compared to related numbers. One of the most effective ways to compare quantitative data is to juxtapose two dimensions on a Cartesian coordinate system, or x-y plane. This works well because the eye immediately grasps the line length and 2D position, while areas like boxes of different sizes or slices of a pie are harder to differentiate.

For example, the larger circle has 16 times the area of the smaller circle. Most viewers will underestimate the difference. Curved edges and the lack of a baseline lead to a consistent underestimation of the actual percentages (Cleveland & McGill, 1984).

Certain relationships lend themselves to specific chart types better than others. Bar charts have many applications: they first and foremost encourage focus and comparison of individual values. The length of the bar encodes the values in a highly readable way, as long as the bars start at zero.

Bars are a means of encoding percentages, too, avoiding the visual distortion pie charts produce. A stacked bar chart shows part-of-the-whole as individual segments of a single bar.

Bars that represent frequencies of a single random variable are called histograms and show the distribution of the variable over the range of possible values.

Bars organized around a zero axis effectively show deviations from a target, such as the break-even point in business or net debt/income.

Points emphasize individual values, rather than the shape of those values. They can encode values along two quantitative scales simultaneously (correlation) in a scatter plot, and may also replace bars if the scale does not include zero.

Lines show the progression of data over time and are therefore ideal for time series data, displaying the shape, trends, and patterns. Plotting two lines side by side yields a comparison of two series over time. Lines may complement histograms with a continuous frequency polygon or add a regression line to a scatter plot.

Small multiples show many similar charts of different data in a very small space, so that they can be seen and compared all at once. By keeping the scale consistent across all the miniature charts, the condensed information is still very readable.

Box plots show the distribution of data, similar to histograms, with the additional features like the median, quantiles, and whiskers, all helping the user see the skew and outliers of the distribution more clearly.

Remove Distraction and “Chart Junk”

Anything that does not contribute to the meaning of the data distracts communication. Remove things like bright colors and fancy backgrounds. Subdue necessary grid lines and labels.

It is a good basic practice to use relatively soft colors in graphs, such as lowly saturated, natural colors found in nature, reserving the use of bright, dark, and highly saturated colors for those occasions when you need to make something stand out.

Use only 5 to 10 major tick marks on an axis to avoid clutter.

Hide distracting data series, while keeping the accessible on demand, use filters/slicers to allow users to focus or zoom out.

Readability

• Highlight specific data with contrasting borders, thicker lines, or larger point sizes.
• Label lines directly at their endpoints instead of using a distant legend.
• Position variables you want to compare next to each other in the layout.
• Include notes directly on the graph to describe specific events or provide instructions on how to interpret complex views.
• Arrange legend labels in the same order as the data bars/lines they represent.
• On very large graphs, place scales on both sides to help the eye identify values accurately.
• Let your axis scale to extend slightly below the lowest and above the highest value.
• For scales involving positive and negative numbers, position the axis line at zero.

Sources

• Source: Few, S. (2005). Effectively Communicating Numbers: Selecting the Best Means and Manner of Display, Perceptual Edge.