Conundrum: Correlation vs Causation (are they really the same?)

Sometimes we often get awestruck by some pieces in random news portals regarding random studies or researches like “A study in XXX area between YYY participants reveals that those who do AAA activity are more likely/less likely suffer from BBB etcetera.”

Sometimes some news may come to us surprisingly and we tend to believe that clickbait headline. For example, a few days ago, I came upon a piece of news on a random news portal that a study revealed that in a certain sample population (I forgot the area name, most probably a state in the US), those who smoke have less likely chance to suffer from Coronavirus. This specific news got huge traction and shares in social media.

These kinds of study results may show the accurate data but we often tend to misinterpret those because of getting confused between two statistical standpoints, and they are,

Correlation and Causation

Don’t worry, I am not a statistics expert neither. I myself did tend to think and interpret similar random news or data just like some of you people. I only came to know the simple difference between them when I was studying “Customer Analytics” from “Wharton School by University of Pennsylvania” last month via online and then my perception changed a bit.

Let’s break it down to you,

Correlation is a kind of a concept when two or maybe more events are correlated with each other. These events even can be mutually exclusive. Correlation is mainly defined by the relationship between two variables (or maybe sometimes more). Example: price and sales.

Where causation means one variable creating an effect in another variable. But we often do mistake in interpreting them when we mistake a correlated event with them having cause and effect relationships. 

Still haven’t understood? Let me give you a more simple example.

Say, in a random city of Great Britain, a random statistical research may reveal that in a season when the sale number of Ice-cream increases, the number of death of people by drowning increases too. In another season, when the sales of Ice-cream spikes down, the number of death of people by drowning also gets reduced.

So, what can we conclude? These two events are correlated. But we surely have the commonsense that, eating more ice-cream or increased sales of ice-cream is not the underlying reason for more death tolls by drowning.

For causation, three requirements are needed to be present in multiple instances or hypothesis.

Let's take two events X and Y. The requirements which are needed to be fulfilled are,

1. Correlation

(Evidence of association between X and Y)

2. Temporal antecedence

(X must of occur before Y)

3. No third factor driving both

(Other possible factors or external effects should not be present between these two X and Y events)

Ultimately, if these three factors are not fulfilled altogether it will be not a causation.

So what can we conclude, we can conclude that some interesting news regarding any uncommon research or bizarre factors may ‘correlate’, but they certainly don’t have causation or simply put ‘cause-and-effect relationships’

Food for thought

Let’s look at the picture closely and relate it with the theory of correlation and causation. Also see to that, does the causation fulfills those three requirements or not. I do believe that certainly you will be able to come up with more examples.

Big companies often systematically perform causal research based on via these two theories. Sometimes they systematically change their pricing, sales volume, offers, discounts, etc to regain insights. Some tech companies also tweak their different dimensions and metrics (for example their apps’ UI/UX, landing page of the website, etc) to generate more insights and do A/B testing. For example, different types of landing pages will generate different volumes of CTR (click-through-rate). So this information and causal research help them to pinpoint some of their decisions. I would love to write about A/B testing and their practical implementation process, industry practices in global tech companies another day.

Ultimately these every sophisticated test come to the basic understanding of correlation and causation.

Thank you. Feel free to reach out for any queries. Besides, I would love to know your valuable inputs or insights if you want to correct or add something.