Subscribe to receive blog updates

Search form



Sep 16, 2015

The Secret of the Dumbbell Chart!

Welcome to our first guest blog post! This user, Janet Wesner, is a insurance industry professional and dataviz enthusiast. Recently she created this data visualization and it created quite a buzz on Reddit! It quickly rose to the top of our Gallery as a result. We asked her to talk about the creation of the dashboard in detail and this is the result! Enjoy!

Out of all the visualizations I created for Slemma over the past few months, I never would have guessed that the Male and Female Anatomical Difference visualization was the one that would make the biggest splash. Since it did, I feel obligated to make my first blog post about this visualization. Also, the dataset is actually small enough to fully include in a blog post. The top chart is what I’ll focus on as the bottom is much more straightforward.

General Concept:

The dumbbell chart is created by overlaying a bubble plot on top of a bar chart. 

 Bubble Plot:

a.     Bubbles = [ID]
Each bubble is a unique body part and gender combination so I needed some identifying column that captured both variables. The ID column concatenates body part and gender and adds “: ” in between.
b.     Axis Y = [y1]
To create the vertical dimension, I ordered the body parts (from Head to Foot) and indexed them from 1 to 15. I didn’t end up using column y2 but left it in, in case I wanted the reverse the order. I didn’t want to worry about going back to the data, which can be time consuming.
c.     Axis X = [Percentages of Total Body Weight]
This is where the bubbles fall horizontally in the plot.
d.     Color = [Gender Num]
Because Slemma requires Color to be a numerical field, I created the Gender Num column as a numeric identifier for gender. 

Bar Chart:

a.     Metric = [val]
The trick here is that I want two bars – The min bar is calculated as the min of the female or male value and the max bar is calculated as the max of the female or male value.
b.     Bar = [bar]
Since the point of this plot is to show the connector between the male and female markers in the bubble plot, I colored the min bar white and the max bar gray.
c.     Group by = [Segment]
The ordering of the segments needs to match the bubble plot.

 Putting it all together, it’s important to keep in mind that there’s a user mouse-over function that pulls up additional chart information. That means that for useful information to show, the bubble chart should be on top. The Female/Male legend was manually created using the circle shape and text. To add some more style, I created colored labels that align with the gender has the higher value of body weight percentage. I go back and forth on if I think this last part was a good idea.

What I Think I did Well

Creative visualization: It’s not often that I get to toot my own horn so I’m going to go all out here. Not to say that this design is a completely new concept (see here and here), but the “dumbbell dot plot” is at least not widely used and I haven’t seen it used for this particular dataset. Also, it took some resourcefulness to figure out how to create this plot because it’s not one of the built-in options in Slemma.
Color coding: While I understand that the coloring I used is very stereotypical (red for females and blue for males), I think a good visualization should consider cultural context. If the purpose of my visualization was to convey the information about body part differences between males and females, then using the color that is most commonly associated with the respective genders make sense. For those who do have a problem with the color choice, just remember that pink only became associated with females because of the feminist movement and was previously thought to be a masculine color.

What I Would Change for Next Time

Focused more on the data: My guess is that the popularity of this data visualization had more to do with the title and data behind the visualization than the actual visualization. While I mainly wanted to see if I could create a “dumbbell dot plot” in Slemma, I didn’t think too much of the data and irresponsibly released it on the internet without any caveats on the data (I have since added some more background information). In hindsight, knowing that people are more interested in the data here, I should have given more thought to data selection and considered things like how the obesity epidemic could have affected results.
Make it colorblind friendly: Though I’m sticking to my color choice, I think it would help if I made the female and male markers different shapes (choose two of circle, square, triangle, star, diamond, etc.). 

No comments :

Post a Comment

Share a post