In my last post, I gave you a couple of triads and dyads to play with. Part of how I analyze story data is to see if there are statistical differences based on how respondents answer multi-choice questions (e.g., How common is this story?) and demographics (e.g., Primary geographic region where you work). In the example shared here, I am working with the Employee Satisfaction story set again. Normally, I need several hundred stories before I run statistics (for validity purposes). This is a public data set that has only 70 stories, but will give you the concept.
Since we do not see normal bell-shaped curves (or a Gaussian distribution) produced with our triad or dyad responses, we need to use non-parametric statistics. I am using the Kruskal-Wallis H Test to compare responses to multi-choice and demographic questions. With any finding that shows a statistically significant difference at the 0.05 level or less, I run a Post Hoc Test called Fisher’s Least Significant Difference to pinpoint which of the answers are different than others. When I ran this test on the Employee Satisfaction story data, only one mulit-choice question showed a significant difference with one dyad.
Now we can look at this dyad to see the difference visually.
Use of statistics like this, help to prevent your eyes from seeing differences where none exist. Go back to the previous post. Go to the third story point on the Tableau embed and try it yourself.