Got Likert data? Put the Neutrals off to one side - Data Revelations

• Read original article here

Got Likert data? Put the Neutrals off to one side
Got Likert data? Put the Neutrals off to one side
November 3, 2019
Overview
Anyone that follows this blog or visits datarevelations.com to read articles on visualizing survey data know that I spend a lot of time thinking about how to present Likert scale data and what to do with neutral responses .
I remain a stalwart supporter of some type of divergent stacked bar chart, but I now advocate placing the neutral values (the mid-point responses when dealing with odd vs even Likert scales) off to one side.
In this post I’ll present the two views I recommend, then get into the specifics on how to build the two views using Tableau.
Note: For all the examples shown below we use a five-point Likert scale with the following sentiment assignments:
1: Very Negative
5: Very Positive
Scenario and solutions
Here’s the results for the question “For the following statements, indicate the degree to which you agree or disagree.”
Figure 1 — Likert scale survey data
If I did not know my audience and did not have an opportunity to show them what I think would be a richer, albeit less intuitive solution, I would go with the approach shown below.
Figure 2 — Simple divergent stacked bar chart with neutrals on the side.  
Hovering over a bar provides details, as shown here.
Figure 3 — Hovering over a bar shows details.
If I knew my audience could “handle it” or if I had time to show them the benefits, I would show all levels of sentiment for positive and negative, as shown below.
Figure 4 — Showing four levels of negative and positive sentiment with the stronger responses hugging the baseline.
So, why the funky ordering going from disagree to strongly disagree (oranges), then strongly agree followed by agree (blues)?
Placing the stronger responses along the baseline makes them much easier to compare. Most people want to see overall positive or just compare the strong positives (the darker blues). I can’t imagine wanting to compare just the light blues.
When you compare just dark blues some interesting things pop out. “Can Play Jazz” is ranked sixth overall, but if we just look at Strongly agree…
Figure 5 — Comparing responses for people who strongly agree.
… we see that this item is tied for first along with “Makes good coffee.”
Update: I showed this post to my friend and Likert scale sparring partner, Daniel Zvinca, who suggested another approach that I include in the embedded Tableau workbook. Here I separate the positives, neutrals, and negatives into three separate columns. Note that this view is very easy to render in Tableau.
Figure 5a — Three-column approach.
Building the simpler view
Now that we’ve seen the two divergent approaches, how do we build them in Tableau?
Let’s look under the hood at the simpler version.
Figure 6 — A dual axis chart (1) for the negative and positive values and a second chart (2) for the neutrals.
Here we see a dual axis chart (1) that displays the orange negatives and the blue positives, and a second chart (2) that displays the neutrals.  I’ll explain why we use a dual axis chart in a moment.
The field [% Positive] is defined as
  SUM(
     IF [Value]>=4 then 1
     else 0
     END)/
  SUM([Number of Records])
This translates as “if the response is 4 or 5 make this a 1, add all the 1s, then divide by the number of people who responded.”
The field [% Negative] is defined as
SUM(
  IF [Value]=4 then 1
  ELSEIF [Value]