Fiddling with data

Photo by Lucia Macedo on Unsplash

I am always looking for new data visualizations that help stakeholders see something they might not understand with numbers or words. We can take the results from a Slider question and present it in its histogram format and most people will be able to interpret the responses overall.

A histogram of a Slider question results

We can tell that the vast majority of respondents shared experiences that were examples of ‘total despair’. One way I like to analyze the data further is to use non-parametric statistics to find differences by factors such as demographics or responses to Multi-Choice Questions (MCQs). I talked about this a while ago.

I ran a Kruskal-Wallis H Test looking at different factors. With the Emotional Tone question, I re-coded into three choices, i.e., positive, mixed, or negative, and ran this test. The results had us reject the Null Hypothesis. There were indeed differences based on the emotional tone of story someone shared. When I found this difference, I wanted the stakeholders who are making sense of these results to be able to grasp those differences and not just take my word. One way to do with is with a Box Plot initially conceived by John Tukey and published in a 1977 book.

A Box Plot of a Slider showing the median

This Box Plot is pretty clear in demonstrating the differences in emotional tone and some people will see those differences. It is a bit technical looking with it outliers and the case numbers. I wanted to come up with another way that was more engaging, drawing in those who were making sense of the data.

Building off of Tukey’s Box Plot, Jerry Hintze and Ray Nelson came up with the Violin Plot in 1998. A Violin Plot builds a kernel density plot to project a shape along the X axis. Wikipedia has a great article which explains it further. And here is a Violin Plot of the same data that you saw in the Box Plot.

A Violin Plot of Slider question results showing the mean

It is the kernel density plot that highlights the differences and makes it easier for the stakeholders to comprehend.

Violin Plots are pretty easy to create if you have any capacity with “R”.  You use ggplot2 with the geom_violin function.

There is good news if you don’t use “R”. I’ve been working with KDV Decisions to develop an “R” suite of visualization tools that will work with story data from the platform. Talk to me if you want to know more about these tools.

Leave a Reply