Can correlation show causality or misconception?
Most of the time, we tend to seek for understanding and explanation of the events that occurred in our life. For example,will drink and drive cause car accident? By using correlation and causality to look at the event, this helps us to make sense and logic about our world. Correlation tell us about the relationship between the events. For example, a person who likes chocolates generally likes ice cream more than crisp. whereas causality tell us the cause and effects of the events. For example, a person will drunk due to the effect of alcohol. However, do we apply correlation in finding what causes the event to happen in our daily life or have we misused it??
There are a lot of misconceptions about correlation implies causality. Causality is often misused by many people which due to the pattern of causality in mind. Our mind is subconsciously establish links between many pieces of information and tries to construct a relationship between the information and to make a conclusion or assumes that a relationship is exists. One of the common errors is that information from media sources always misinterpreted by many people. Based on the example above, drink and drive will cause car accident. So, people who wanted to avoid the accident tend not to drink any alcohol drinks. They fail to understand that, just because the result show a correlation , there is no proof of an underlying causality. Indeed, drink and drive can cause car accident but not necessary all people who drink and drive will cause car accident. Drink and Drive will only increase the probability of car accident. Furthermore, there are other factors such as tiredness, lost control of the car or car breakdown which also can cause car accident.
One of the common errors of confusion between correlation and causation often can be found in scientific and health studies. One of the obvious example is a British newspaper that reported a survey about the behaviour of teenagers and whether their parents smoked. As a result of the survey, the newspaper reported that the children whose parents smoked were more likely to indulge in delinquent behaviour. Indeed this result of the survey show a correlation between children’s behaviour and whether their parents smoked. However, the paper printed out the headline , which ” Parental smoking causes children to misbehave” There are a few possible explanation with this assumption. It could be parents smoked because of having stress in handling children misbehaviour or it could social class, where the lower classes are more chances to smoke as due to the stress of poverty and more likely to have delinquent children. Unfortunately, our intuition can lead us to this confusion in distinguishing between correlation and causation. Thus, parental smoking and children misbehave may correlated with poverty, but it may also have no direct link between them.
(Source http://www.criticalthinking.org.uk/smokingparents/)
On the other hand, just like the concept of validity is related to reliability,where “just because the results showed are reliable, doesn’t mean that it is valid, though valid measurement must be reliable”, the concept of causality is also related to correlation, where if one event causes another, then they are most certainly correlated. But just because two things occur together, doesn’t mean that it showed the causality between the events, even if it seems to make sense. For example, smoking a cigarette not only correlates with lung cancer, but it actually causes it. However, if smoking is correlate with alcoholism, but we do not know whether alcoholism got a real effect on lung cancer or not. All we know is that, a person who is alcoholism and smokes cigarettes has higher chances of getting lung cancer. Thus, It is true that cigarette causes lung cancer, but it is hard to establish that alcoholism can cause lung cancer.
In conclusion, indeed causality event can be correlated, but does’t mean that correlation shows causality of the events.Correlation increase the chances of causality relationship, but do not implies the causation of the event. These correlations simply show a relationship and tell us nothing about the explanation of that relationship.
Reference :
http://www.experiment-resources.com/correlation-and-causation.html
http://www.johnmyleswhite.com/notebook/2010/10/01/three-quarter-truths-correlation-is-not-causation/
Posted on March 9, 2012, in Uncategorized. Bookmark the permalink. 5 Comments.
psychrno 
I think that although correlation gives us a good background on the direction, degree and form of a relationship between two variables, we caanot assume causation from just using correlation. Correlation does not give us any information on what variable causes the other it just states that when a +ve correlation is evident, the two variable go in the same direction i.e they both increase. A weak correltion can be shown when the two variables show an inverse relationship so as one increases, the other decreases. The degree of the relationship tells us that a perfect correlation is at 1.00 and demonstrates a strong relationship and anyhting below this is not perfect. Another factor which correlation does not take into account is the third variable so for example alcoholism and antiocial behaviour correlate but we cant say which causes which and it could also be another thing involved such as stress levels which may cause alcoholim. or family background in these areas.
I totally agree the points raised in this blog entry.
I’d like to highlight some of the reasons behind our assumption of causation rather than correlation:
1. The Media
– as you mentioned above sometimes newspapers misreport correlations as causations and this may have been a result of…
– the reporters lack of knowledge (maybe some of the reporters do not understand the results of the report themselves – and this may be increasing because scientific reporters are declining).. an obvious and ineffective way to solve this is for the researcher or a professional in field approves the article before printing. However this is very difficult as all reporters work on a quick and strict deadline; there just wouldn’t be enough time.
– to simply sell more newspapers. As it’s their jobs on the line.
2. Questioning reported results.
– Usually those who do not specialise in the field of the research being reported (in the media) tend not to question the results. It may be due to lack or critical analysis skills.
– Another problem is that we tend to believe that there is causation when reported in scientific studies. We tend not to question their results or their method simply because they are professional researchers – we assume they know what they are talking about. But we’ve got to remember that when we are reading research.. all psychological research papers have at least one flaw. So therefore if only one paper states there’s a causation.. can we really believe it? We should question the methodology of the study.. and I guess we can only agree that the finding was indeed a causation and not a correlation if the study was replicated using many different methods. However, we can only agree that is indeed a correlation until it has been refuted.
I think you have really argued the points well 🙂
I agree with your point, correlation does not necessarily signify causation. However, as you said above, correlation does increase the chances of there being a causal relationship.
However, we must ask ourselves if we can truly find a causal relationship? as ‘amybe91’ said above, we must all take into account extraneous factors and a third variable may be contributing to the effect that we find. So, (in using your example above about smoking and lung-cancer) as you rightly said, smoking does cause lung cancer , however this is not always the case. Some people smoke from the age of 13 and live a long, healthy life whereas others will smoke for a few months and will unfortunately get lung cancer. Samet, Humble and Pathak (1986) found that an individual’s family background was also a key factor in the development of respiratory diseases and lung cancer and this could possibly be an extraneous variable contributing to the relationship between lung-cancer and smoking.
Overall, I think you make a really clear and concise argument about how correlation does not signify causation in all cases.
🙂
Pingback: Week 8 comments for Alex :) « afshinpsychology
Pingback: 14.O3.2O12 – COMMENTS! « nim2152