Visualizing survey responses
Often surveys contain questions involving a Likert scale or rating scale (strongly agree, agree, disagree, strongly disagree) . Visualizing this type of data can be problematic because there are many different ways of grouping and analyzing responses. A common approach to dealing with this issue it to collapse the results into two groups, e.g. strongly agree or agree v. disagree or strongly disagree. Another approach is to present the results in the form of a stacked bar chart. Both of these approaches have limitations. Grouping the data results in the loss of potentially valuable information, e.g. one may want to know which question had the most respondents that strongly agreed. The stacked bar chart, while valuable for comparing cumulative values, makes comparison of the individual components difficult.
The following is example of a tool that facilitates both comparison of individual components as well as cumulative values. The tool also illustrates an effective approach to analyzing data with multiple categorical variables, which is also common among surveys. By representing the categorical variables within columns rather than across a row, we maintain the ability to filter and sort the category’s values. The tool also uses layering as a means of creating reference lines. In sum, with this approach a tremendous amount of comparisons are available within a few seconds. In previous posts I have presented a similar tool, VBA to assist in small multiple creation, and layering. The actual graphs in this example are a bit odd for two reasons: (1) I neglected to fix my axis, and (2) the random data I generated does not add up to 1.
Note: workbook provided upon request.
Hi there. So I guess this is effectively a trellis display?
Just a quick thought…if for some reason you only had room for say a stacked bar chart on a dashboard, you could always add a dropbox where the user selects whichever series they want to appear closest to the axis, allowing them to compare each series one at a time. Not ideal, but might work in some circumstances.
Or you could consider the approach over at http://www.excelhero.com/blog/2010/03/animated-stacked-chart.html
Yeah, I assume this could be considered a trellis display.
Dynamically selecting the variable next to the axis could be very effective, particularly is space was limited. It may work best when the elements lack a natural ordering, e.g. ethnicity. In cases where ordering is relevant, such as this survey example, re-ordering the elements could be confusing. Placing disagree first would result in the following order: disagree, strongly agree, agree, and strongly disagree. Perhaps strongly agree could be moved to the front with disagree, but reshuffling multiple components would also be confusing.
Assuming space was limited I would prefer to have two graphs. (1) The fixed cumulative stacked bar with its natural ordering, and (2) a single competent of the stacked bar that could be selected dynamically.
Regarding the animated stacked chart. I find the visualization and its dynamic nature to be impressive, but I think it fails in terms facilitating comparison, which is one of the primary goals of any visualization (at least if it used for any serious analysis). At no point can a part be effectively compared to the whole, nor can two parts be effectively compared (e.g. housing v. food). To do these comparisons, the reader must compare an image they are viewing now with one that existed a few seconds ago. Making these comparisons across time is very difficult and inaccurate, and we should strive to put any data we wish to compare within the same page. In sum, a far less flashy tool could better facilitate analysis. At some point, I intend to post an example of a modified version of the chart that retains some of the impressiveness of the original but has functionality that aids in comparisons.
this looks pretty impressive. can you share the workbook?
Thanks. Just sent the workbook.
It looks really good. Any chance of getting a workbook? Thank you in advance.
Best regards,
Jure, Slovenia
Thanks. Sent.
Can you please share the workbook? Very Interested.
Hi,
I’d be interested to get a copy of your workbook if possible.
Thanks
Andrew
Thanks. Sent .
Sent. Thanks.