Correlation vs Causation: What is the Distinction?

Information is a particularly highly effective component of any determination. American statistician W. Edwards Deming as soon as stated, “In God we belief. All others should carry information.” However too usually, information may be misconstrued. One of many greatest confusions throughout any information evaluation revolves round correlation vs. causation.

There are numerous articles that share wild, usually tongue-in-cheek conclusions because of two strongly correlated information units. For instance, Harvard Business Review as soon as checked out examples exhibiting the “risk” that:

  • Spending extra to see sports activities matches reduces your chance to eat high-fructose corn syrup
  • Extra iPhones offered means extra folks die from falling down the steps

These are excessive examples. Though, correlation doesn’t essentially suggest causation, and these examples present the risks of not understanding the distinction between correlation and causation in the true world. In these circumstances, further vetting is required earlier than a correlation can qualify as causation.

What’s the Distinction Between Correlation and Causation?

Let’s begin with the fundamentals. What’s the definition of causation versus correlation?

What’s correlation?

The Australian Bureau of Statistics offers a fantastic definition of correlation:

“[It is] a statistical measure (expressed as a quantity) that describes the dimensions and route of a relationship between two or extra variables.”

In different phrases, a change in a single variable will usually be mirrored by a optimistic or adverse change within the different.

What are the several types of correlations?

  • Optimistic correlation: Variables A and B transfer in the identical route. For instance, as Variable A will increase, so does B.
  • Detrimental correlation: Variables A and B transfer in reverse instructions. For instance, as Variable A will increase, B decreases.
  • No correlation: There is no such thing as a obvious hyperlink between Variables A and B.

The power of the linear relationship between two variables, additionally known as the correlation coefficient, can vary from -1 (adverse correlation) to 1 (optimistic correlation). The nearer the correlation coefficient is to both -1 or 1, the stronger the connection. However, a correlation coefficient of 0 signifies that there isn’t any correlation between these two variables.

Nevertheless, a correlation doesn’t essentially imply the given impartial and dependent variables are linked. Which brings us to causation…

What’s causation?

Also referred to as ‘causality,’ the Australian Bureau of Statistics goes on to outline causation the next manner:

“…one occasion is the results of the prevalence of the opposite occasion; i.e., there’s a causal relationship between the 2 occasions. That is additionally known as trigger and impact.”

In different phrases, does one variable really affect the opposite?

Causation vs. Correlation Examples

Spurious Correlations is an entertaining useful resource that shares examples that present sturdy relationships between variables however that aren’t brought on by each other. Not less than, they shouldn’t be.

Living proof: is consuming margarine behind Maine’s divorce fee?

Chart showing a strong correlation between the divorce rate in Maine and margarine consumption

Supply: (hyperlink to license)

Sticking to meals examples, might cheese be the key gasoline that powers civil engineers of their research?

Chart showing a strong correlation between mozzarella cheese consumption and the number of civil engineering doctorates awarded

Supply: (hyperlink to license)

Each charts present sturdy correlations between dependent and impartial variables. Nevertheless, these are possible basic circumstances of “correlation doesn’t suggest causation.” That’s, until margarine is certainly sensitive topic for {couples} in Maine or there are new ground-breaking results to consuming giant quantities of cheese.

Why is Figuring out the Distinction Between Correlation vs. Causation Essential?

The correlation and causation examples above present the significance of getting the distinction proper is essential.

Avinash Kaushik, Digital Advertising Evangelist at Google, wrote in 2016 about how not understanding the difference can be very problematic. Kaushik highlighted an article from The Economist that asserted that consuming extra ice cream can enhance pupil scores on the PISA studying scale.

“To regular folks (non-Analysts), this graph and article appears legit,” wrote Kaushik. “in spite of everything it is a respected website and it’s a respected staff. Oh, and look there’s a crimson line, what appears like a plausible distribution, and a R-squared!”

However Kaushik desires us to assume a bit more durable concerning the information at hand, and never take issues at face worth.

He factors out that there’s nothing to floor the causation of 1 and the opposite regardless of an affordable correlation. There might seem like a hyperlink connecting IQ to ice cream consumption. Nevertheless, the information doesn’t definitively reveal something apart from that apparent correlation.

Making Daring Claims

In our on a regular basis lives, we’ve entry to extra information than ever earlier than. Selections, opinions, and even enterprise methods can depend upon our capacity to inform the distinction between them.

Kaushik makes use of the instance above to remind folks to be extra skeptical of claims that draw daring conclusions from correlated information factors. He encourages readers to look deeper on the information and keep away from the straightforward selections.

“Our job is to be skeptical, to dig and perceive and poke and prod and to reject the outrageously unsuitable and if it’s not outrageously unsuitable then to determine how proper it is perhaps to be able to make an informed suggestion.” – Avinash Kaushik

Causality vs. correlation can also be a subject that Michael Molnar examines in a Forbes article. Molnar warns that:

“Complicated correlation with causation just isn’t an unknown concern however it’s changing into more and more problematic as information will increase and computer systems get extra highly effective… it will get to the guts of what we all know – or assume we all know – about how the world works.”

It may be tough to infer causation between two variables. Randomized managed experiences and different statistical assessments are sometimes wanted to validate if one variable does, actually, affect one other. Furthermore, whereas correlations may be helpful measures, they’ve limitations. As we noticed within the correlation vs. causation examples above, it’s normally related to measuring a linear relationship.

Getting Correlation vs. Causation Proper

In right now’s data-driven world, being extra skeptical of particular findings earlier than making daring claims, as Kaushik suggests, is important. How can we do that? Additional analysis and, at any time when doable, further testing.

Outdoors components (known as “confounders” or “lurking variables”) can generally come into play for one or two of the variables in a given correlation. For instance, some research discovered a hyperlink between espresso consumption and danger of lung most cancers. Nevertheless, “smoking” has been discovered to be a possible confounding variable within the outcomes, as one meta-analysis of those findings exhibits [1]. As talked about, and as with different key findings, additional analysis might help make clear the context behind correlations.

Testing for causality is difficult. Nevertheless, experimental design might help. That is the place a researcher can check a speculation in a manner the place they will management one variable (the impartial variable) and measure its affect on one other variable (the dependent variable). Most significantly, it might assist them management for doable confounders to keep away from potential bias of their outcomes. For extra details about how experimental design works, this overview by Britannica offers a wonderful introduction.

Method Your Buyer Expertise Analytics with Confidence

At Astute, we assist high manufacturers measure and elevate the shopper expertise with confirmed buyer engagement options, together with Astute VoC, our voice of the shopper answer. Our VoC specialists are additionally there to assist make sense of your information so that you get actionable insights you may really feel assured about.

See how we might help. Request your personalized demo today.

Further assets about correlation vs. causation

Beneath are some nice assets that designate correlation vs trigger and impact.

[1] Galarraga, V., & Boffetta, P. (2016). Espresso Consuming and Danger of Lung Most cancers—A Meta-Evaluation. Most cancers Epidemiology Biomarkers & Prevention, 25(6), 951–957.


Please enter your comment!
Please enter your name here