Corr Studies

Posted on Leave a commentPosted in Calculation Lessons, Today I Learned

Tableau 10.2 came out a couple weeks ago and a small but notable feature included in this version of our product is the Corr() and Window_Corr() functions. These calculations compute a coefficient of correlation based on two variables. I’ve done linear algebra in Tableau before, and it’s often quite complicated and looks like this:   The idea with making Corr a simple function in Tableau is to give users a faster and simpler way to find statistical results. But I was confused. correlation coefficients operate on large groups of numbers, so why would it make sense to aggregate a pair of measures with Corr()? How would the aggregation change based on the layout of a Tableau worksheet? And how could I check the results to make sure I was seeing the correlation coefficient I wanted to see? When doing complex calculations, you often get a seemingly arbitrary number as a result, and some faith is required to trust that the number is answering the question you intended to ask, and not a different question. Below are some findings from my Corr studies: A quick preamble about my math background. I was always pretty good at math growing up, but I […]

It’s Big, It’s Heavy, It’s Wood: When Log Scales Are All Right

Posted on 2 CommentsPosted in Today I Learned

What rolls down stairs Alone or in pairs, Rolls over your neighbor’s dog? What’s great for a snack And fits on your back? It’s log, log, log!   I’m known for tenuous analogies, so I won’t spend too much time defending the connection between the classic song from TV’s Ren & Stimpy and the logarithmic axis on charts. Except to say this: like a heavy log, this brute force tool is rarely the appropriate weapon of choice for representing a numeric range. A couple of weeks ago, Andy Cotgreave explored the notion of a logarithmic axis on one of his Makeover Monday vizzes. In this case he was attempting to deal with the vast difference between soccer (née “football”) salaries between the highest English league and other, lower divisions. His attempt (click to read his original article): As Andy notes in his post, the logarithmic scale doesn’t fix the problem of range in this view: The default scale slope (on the left) makes all the lines appear as if they start from the same point. I tried a log scale which fixes it, but I think log scales are misleading for many viewers, who either don’t understand what they show, […]