Correlations and your relations

When two phenomena evolve in parallel or simultaneously, they are correlated. But what does this observed correlation mean? What should we do and think about it?

 

For example, did you know that in Maine, between 2000 and 2009, margarine consumption per person was almost perfectly correlated with the divorce rate?

 

 

Surprising news, isn't it? But don't question the state of your relationship with your significant other (just yet) – or, worse, the foundations of your joint food purchases. Let's first take a moment to interpret this puzzling finding.

 

Is margarine consumption decreasing BECAUSE fewer couples are getting divorced? So are fallen lovers less likely to turn to margarine to soak up their infinite sadness?

 

Or, conversely, is the divorce rate plummeting BECAUSE margarine consumption is decreasing? So are couples’ fundamental disagreements about which source of saturated fat to include in their diet reduced if they both like... butter? Or maybe peanut butter?

 

Hard to say? This is probably because a correlation:

1. only very rarely reflects a causal link between two variables; and

2. may well be, and is often, the result of chance.

 

Like the very strong correlation between drownings and Nicolas Cage movies.

 

 

Far-fetched associations also exist in the field of health. How do you know if two phenomena are really correlated? What distinguishes a correlation from a causation?

 

If you are struggling to find answers to these questions, console yourself: many of your colleagues are in the same situation, and this is normal. Check out The Running Clinic’s brand-new online course, 1.6 Science in Practice: Applying Research for Clinicians!

 

Taking this course will make you become better clinicians in three complementary ways:

 

1. You will finally be able to identify articles and journals you can trust, among the millions that are listed each year by Doctor Google.

 

2. You will better treat your patients by relying on studies with the highest level of evidence, and the lowest risk of bias.

 

3. You will finally know how to confidently interpret the results of literature reviews to guide your patients, instead of browsing the abstracts and hoping for the best.

 

In other words, you will develop the necessary expertise to interpret all studies about diagnosis and treatment – whether the experimental design is cross-sectional or prospective.

 

"Cross- what?!?"

 

That, too, we explain in the course. In simple terms.

 

We look forward to helping you better interpret science!

 

Note: The figures are from the excellent Spurious Correlations, by Tyler Vigen (2015, Hachette Books), which contains countless other pointless correlations that will leave you clueless.

Jean-Francois Esculier & Olivier Roy-Baillargeon

R&D, The Running Clinic