Tables can be customized to make them easier to read or consistent with other needs. Common ways of customizing include: merging categories, nets, spans, reordering, statistical significance, general formatting, heatmap shading and other conditional formatting, flattening, and changing which statistics are shown.
The table below on the left shows the proportion of people to have selected different age categories in a survey. The table on the right shows the same data, but with some of the categories merged.
A net is a merged category when the original categories are left on the table. In the example below, TOP 2 BOXES is the merge of Love and Like.
A span is a heading that spans across multiple other headings. In the table above, Age Merged appears in a span across the different age categories, with Gender across the sexes.
The default order of categories on most tables is dictated by either alphabetic order or the order that the categories that appeared when the data was collected. The ease of reading tables can be improved by reordering the categories, to reflect either a more natural ordering of the categories, or, pattern in the data (e.g., ordering the categories so the data is sorted from lowest to highest).
Sorting Tables and Visualizations and Diagonalizing Tables and Visualizations describe the principles of ordering data on larger tables.
The tables above each display a single statistic. It is common to create tables showing multiple statistics. As an example, the table below shows Column %, Row %, Count, and Average.
Spans are added to tables automatically when creating banners, or manually when the person creating a table wants to emphasize an aspect of the data.
The way that statistical significance is represented, and computed, is another standard way of customizing tables, with arrows, colors, and cell shading commonly used to display exception tests, as in the example below. Letters are the main ways of communicating statistics significance when using column comparisons
See Introduction to Significance Testing for when and how to use significance testing.
The tables above are formatted with more modern web-style designs. Older software instead generally creates black and white tables designed to minimize the size of the table, so as to save printing costs.
Heatmap shading and other conditional formatting
Conditional formatting can be used to format data conditional upon the values of the data. For example, heatmap shading can be used to emphasize higher and lower values, as in the example below.
Other ways of making the appearance conditional upon the data include:
- Highlighting rows containing particular brands.
- Replacing numbers with symbols or making cells blank when the sample size is too small or the value is below some threshold.
Grids tend to be unhelpful in crosstabs. Consider the table below. It shows attitudes to different brands of colas (the grid) by age. The table reports that it shows row percentage, but it is hard to work out precisely what that means as the columns represent two separate ideas - attitude and age.
The solution to this problem is to flatten the grid. By flattening the grid, we present both dimensions of the grid - in this case, the brand names and the attitudes - in one dimension of the table, as shown below.
Please sign in to leave a comment.