A few months back, we posted a blog on the fun side of apparent causation in statistical relationships … and some gentle warnings as well. Our inspiration primarily flows from a recent visit to a terrific website developed by Tyler Vigen, a Harvard criminology student with a wonderful sense of humor and a sharp eye for spurious correlations. We wholeheartedly recommend a visit to his website.
As we noted last October and are delighted to reiterate, Vigen is quite imaginative at spotting some over-the-top, ridiculous, apparent cause and effect relationships to make his point: “beware the spurious correlation.”
There are classic types of spurious correlations, some quite ridiculous, some not so funny.
Bees, Marital Bliss or Where You Live?
Here’s one from Vigen’s website that brings together two disparate trend lines. They are so ridiculous that it’s easy to see that this apparent relationship is judgmentally so unlikely that it’s laughable.
Here’s a near-perfect correlation between the trend in Apple iPhone units sold and the number of deaths from falling down stairs. An apparent cause/effect relationship? Really?
A direct effect of A causing B: more people checking their smart phone screens when they probably shouldn’t be?
Could it be a lagged effect of B driving A: the notion that increased fear of accidents is driving the desire to have an iPhone handy in case of a fall?
Time for some business-relevant insights?
Is the apparent data relationship simply a reflection of population growth?
So, when the brand health tracking shows significant wave over wave growth in brand funnel metrics, is it the new brand advertising or are there other factors that drove the outcome (such as increased retail shelf presence)?
To minimize falling victim to the spurious correlation, we need to be willing to challenge apparent correlations and offer other business-relevant solutions.
Researchers need to be judges – by challenging apparent correlations and providing business-relevant, realistic insights!