Is taste and political affiliation related?

This project is a dashboard challenge conducted by Maven Analytics. It uses the data from the Great American Coffee Taste Test conducted by coffee expert James Hoffman and Cometeer. The goal was to help the stakeholder open a new coffee shop.

While developing this report, I learned about the Chi-Square statistical test to compare the different political group results to see if there was significant evidence to show there is no statistical difference between the varying percentages.

To create the Statistic Test table displayed on the report, page Bias, I had to create a virtual table and link it to the Participant_Info table. That was a good puzzle to solve. I soon found out that the DAX function SUMMARIZE does not work on virtual tables to group by Coffee types. This is where I learned the DAX functions GROUPBY and CURRENTGROUP(). Once the sum of each Political Parties for each coffee type was accomplished the table was quickly linked and you can see the results in my report.

During this challenge there was talk about using Power Query to unpivot the survey questions. At first, I questioned whether this is the most optimized way of doing the task at hand. However, after experimentation, I soon realized it was the easiest method. It also allowed for easier corrections to the table than the usual merge method used for creating look-up tables. I minimized the amount of rows in the unpivoted table by filtering out the answers with FASLE in them. So a compromise was reached between optimization and ease of doing the task. See: Data Model in pictures below.

Another interesting thing that occurred in this challenge is the concept of how a the results of a taste test can be influenced by political affiliation. This is discussed briefly on the BIAS report page. What made me consider this is when I saw how the dataset had a large number of Democrats compared to Republicans. It would still be interesting to find out how the samples were mailed out.