After 5 years at Microsoft I decided to move to Tableau Software to work on an area I love; data visualization. Tableau is an interesting piece of software. It is easy to learn and generally very easy to use. Another nice thing about Tableau is there’s a way to do everything even if it is not supported via simple point-and-click out of the box. Table calculations, using multiple measures… you’re only bounded by your imagination. I came across a lot of people asking about Trellis charts and even more people on how to do linear regression in Tableau.
One thing to clarify, Tableau already does trend lines and even time series forecasting, not to mention providing you with a handful of statistical test results indicating how well these fit the actual data. But sometimes you may want to access those values e.g. use the equation for the trend line in a calculated field or filter out bad fits by using regression coefficient. Unfortunately model parameters are not readily available in Tableau for these tasks.
I put together the example you can see below and while I can do a lengthy description of how it is put together, it is probably best if you download and take a look at it yourself. Workbook and data are published on Tableau Public. In two clicks you can start exploring and modifying the workbook.
In this particular example, I’m plotting 29 different Seattle neighborhoods to see how square footage of the houses relate to the price. Dealing with 29 different charts isn’t easy.
Using color makes discovering patterns and outliers so much easier. So I assigned red to indicate inverse proportionality while green indicates positive correlation. This way even in a list of hundreds of plots I can very easily detect regions that deviate from the rest of the set in terms of the relationship between these two variables. Enjoy! Questions/comments welcome.