Originally Posted by Neel

I'm currently working on a project which generates a price sensitive forecast for a major cruise line company. For my portion of the project, we wanted to pick 5 different key variables, segment the data by every combination of those variables (all of them, all combinations of 4 of them, etc.), and see how the results turned out for each one. I accomplished this by loading all of the possible segmentation levels into a table in oracle which our statistical package then interpreted (replacing variables with nulls, thus making segmentation by said variables irrelevant).

To construct the table, I first created a row where all were marked as "Y." The only ones I toggle are in the middle (switch from white to green). I then copied that one row and swapped the first cell that I want toggled (making it white). I then copied both of those rows, pasted them back, and swapped the second column for that new block (switching it to white). I repeated this process for each variable I wanted swapped, doubling the size of the table each time. The result is a cool fractal-ish ("ish" might not be necessary if this is actually considered a fractal) table that includes every possible combination of my variables in a format that I can use and with very little work required on my part.

Here is a pic:

For fun, I also tried it with three possible states for each variable (1,2,3 rather than Y/N) and used conditional formatting to assign colors based on the color of each cell (same thing that I did here, but with an additional color). It also looked pretty sweet.