Correlation and Causation

It’s well-acknowledged that correlation does not equal causation, however, because there is such a high “correlation” between correlation and causation, people tend to confuse the two frequently and substantially.

Therefore, it is highly necessary to think deep of these concepts and be more clear-headed in coping with problems in life and work, so you don’t have to live in a constant anxious status, getting hit by untoward mishaps, feeling unfortunate and helpless all the time. For instance, you may have met someone who abstained him/herself from any drinking and smoking, sacrificing the relaxation and recreation she/he used to enjoy, whereas eventually still died from cancer. They blindly regarded high correlation between alcohol /tobacco and cancer as causation, so when things turned out in an opposite way, they feel very upset and dismayed.

Attempting to tackle this confusion, I first want to clarify what causation is. It’s straightforward thusly:  A (cause) leads to B (effect or outcome), one-directional and for certain. One directional means it’s A leads to B, not B leads to A; for certain means when A appears, B will be generated with 100% certainty, no exception. So any bogus axiom can be overturned if one exceptional case is observed. That’s the essence of scientific approach in establishing a causal-effect rule. One example is the gravity law – if you hoist a matter to a higher level and let it go, it will fall to the earth with the position and speed calculable per Newton’s equation; another classic example is when the patients are injected analgesic medications before operations, they will be sure sedated for conducting the surgery; third example, before it is entirely eliminated from the earth, infection of variola virus surely leads to smallpox disease in human beings. So height -> fall, narcotic drugs -> anesthesia, variola -> smallpox, they are well-founded causations and controlled by the law of universe fundamentally.

Now think about correlation, the word itself speaks clearly that it only means there are high coincidence rate between A and B. If A causes B, the correlation between A and B is one, so causation leads to correlation but correlation doesn’t guarantee it’s a causation relationship. I summarized into three categories to explain high correlations not companied by causation relationships –  “Association”, Reverse Causation” and “Overcoming Threshold”.

Association can be illustrated with the famous Pavlo’s dog experiment. When the condition between ringing the bell and feeding dogs food is established and augmented for a sufficient time period. The salivation is highly correlated to bell ringing. However, you can not attribute ringing bell as the cause of salvation, it’s a man-made association to food-giving, which is the true cause of secretion of saliva by dogs.

Reverse causation can also be illustrated by an example. There was an article published stating that the longer a patient is hospitalized, the worse his/her health is. The paper is rife with solid statistical evidence, however, you can easily refute the whole thesis because the author confused the simple fact that it is the worse health situation caused a longer staying in the hospital, not vice versa.

By Threshold Overcoming, I am referring to a scenario such as boiling the water,  constant heating can lead to boiling water, but if the flame is weak, no matter how long a time is given, the water will never hit the temperature to be boiled up. It fits almost perfectly into causation relationship except to meet the threshold. In real life, we are exposed to radiation all the time, but the quantity is extremely low that we don’t need to worry about carcinogen effect by sunrays.

Clearing the thinking path by defining causation and other causes of correlation, now we have to deal with most confusing part – I call it quasi-causation phenomenon, which is clearly one-directional from A to B with high probability but not 100% certainty. Such examples are inundant in our daily life such as whether drinking or smoking will lead to cancer, whether texting during driving will lead to car accident etc.

Statistical evidence is overwhelmingly compelling that there is a positive correlation between smoking and cancer, so does a long driving meters related to road accident rate. Even strict cohort studies are conducted, returning the same conclusion – high correlation between A and B. But, A is not the cause of B according to the definition of causation, as long as there is one exceptional case, A is not the cause of B because A -> B is falsifiable.

You can not accuse driving as the cause of car accidents in a similar way that you cannot blame your parents giving birth to you so rending you vulnerable to cancer death. So we exclude driving or living in this world itself is the cause.

Then is “high speed” the culprit of road death? high-speed is always highly correlated to the accident, but the answer is no again. We dwell on the earth which is orbiting with a much higher speed but we are sound safe; furthermore, the airplane is so far the safest transportation vehicle with the highest speed, the case is rested.

So, lastly, is it the reckless behavior such as texting during driving the culprit? This looks very like the prime suspect. But hold on, take a step back and think through, does texting during driving really causes car accidents? Does smoking really cause cancer? When you take into consideration that there are numerous cases when texting didn’t incur car accidents,  or, a more specific example in the latter case, smoking to his eighties, Winston Churchill didn’t die of cancer.

Let’s think about a rare occupation – acrobat. Don’t you feel it is a reckless, highly dangerous behavior to ride on a rope in the air? What’s the confidence gap between you and a skillful acrobat to assume this stint? If you can train yourself to be able to handle such delicate activity, you can do it with ease as walking on street. Similarly, if you can, or technology makes it possible for us to text with enough attention bandwidth to monitor the traffic situation simultaneously, texting during driving is not the cause of car accidents and hence can be done without any problems.

Unfortunately, as of now, texting, acrobating, smoking or drinking – so many things fall under this unknown or non-handlable territory of our knowledge and ability. Being a cautious, risk-averse person, you can always choose to stay far away from such activities, until one day, scientists are able to crack the codes and figure out whether smoking is related to cancer via “association”, “reverse causation” or “threshold overcoming”.  Prior to the advent of this day, we’d have to weigh between the abhor of taking risks and the benefits from risk behaviors such as smoking, drinking or texting when driving.






Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s