Vizible Difference

There has been a calculation that is needed in market basket and correlation analysis. There is a KB article about the topic (by self-join) and I wrote a post about it before (by custom SQL).

Recently Łukasz Majewski has provided an ingenious solution using data densification. I am totally stumped. I am going to applying it to the same superstore example that I used before, and as a result, to document the details.

The objective of the analysis is to count customers who bought in both A and B subcategories of products.

The main steps are:

1.Duplicate Sub-Category field: Sub-Category (Copy). Drag them into columns and rows respectively. We see 17 data marks in the view.

2.Create Index (=Index()) and drag it into Details shelf. Set it to compute using Table Across or Subcategory.

A little question about Markov Chain came up lately. That pushed me to have a closer look at what it does. I had a vague idea about it before. This time I understand it better.

I saw a question in StackOverflow where people seems favoring a solution in SQL. I found it pretty easy to calculate in Tableau and with better visualization.

So, the matrix includes the probabilities of a group of entities transitioning from one state to the next state.