Charts

From Market Research
Jump to: navigation, search

Our brains are much more efficient at processing images than at processing numbers. By presenting tables as charts we can make it much easier for people to interpret findings. Plotting of Summary Tables using pie charts and column charts is straightforward. Unfortunately, with more complex tables, it is not the case that any chart is better than a table. Our brains are only good at interpreting images when the images are structured in particular ways.

Contents

Use columns for comparisons and lines for trends

The two charts below show the same data. When we look at the plot on the left our brain automatically tries to compare the height of the columns. If we look hard enough and read the comment we can see a trend. However, when we look at the plot on the right the trend is so clear it does not need to be described. The general conclusion is to use columns when wanting to emphasize and compare individual results and lines when wanting to emphasize and compare trends.

LinesForTrends.PNG

Minimize difficult comparisons by using grid layouts

The table below is large and complicated. It is not straightforward to examine it for conclusions.

ImageGrid.PNG

A 'standard' way to plot such data is as a column chart, as shown below. While prettier than the table, it is actually less useful. Consider trying to work out which brand is most strongly associated with being Sleepy. On the table we can quickly read down the appropriate column. In the chart below it is much harder to work out.

ColumnClustered.png

The solution to this charting problem, and to many charting problems, is to create a structure that is more grid-like, as shown below. The reason that this chart is better is that it replaces numbers with bars, without sacrificing the ability to read by rows or columns. Thus, we can quickly scan this chart to identify that, for example, Diet Pepsi scores the highest on Sleepy.

ColumnBars.png

This same basic principle is useful for time series data as well. Consider the default line chart created by Excel, shown below. It is very different to disentangle the meaning in this data.

Lines.png

We can get a small improvement by replacing the legend with individual labels on each line (this makes the comparisons easier because the reader does not have to continually work out which line to use).

Lines2.png

However, the chart below, which again adopts a grid like layout, and also orders the results, is substantially easier to interpret.

Lines3.png.

Use color and redundancy to emphasize

In a traditional chart is used to indicate different series. For example, in this chart blue is used for feminine and health orange for health-consciousness. This use of color is unhelpful. Our brains infer the colors as indicating that the columns are 'different' in some way, which makes it harder for us to accurately compare the columns and thus color in this context completely undermines the purpose of the chart.

ColumnBars.png

A better use of color is in the chart below. Color is still being used but it is being used in a different way. The intensity of the color is proportional to the length of the bars. This redundant use of information helps our brains (see Ordering for a further improvement of this chart).

OrangeImage.png

The next chart improves on the earlier time series chart. Note that color is used here to emphasize interesting results. In this case, it is used as a way of communicating the results of statistical testing.

Line4.png

Software

Most statistical packages can do the types of plots shown on this page (e.g., SPSS, R, SAS). They can also be created with a bit of work in Excel (e.g., by creating multiple charts and lining them up, or, by buying add-ins). The simpler plots on this page were created in Excel. The 'grid'-style plots were created in Q, which is the only specialist survey-analysis program that implements these styles of plots.

Personal tools
Namespaces

Variants
Actions
Navigation
Categories
Toolbox