For this weeks blog, I have decided to look at the topic of correlation and causation.
The title of this post refers to the latin phrase meaning “after this, therefore because of this”. Post hoc reasoning is a logical fallacy that states that because event A happens before event B, A must have been the cause of B. In science, this type of fallacy is often the result of making a post hoc error as to the causes of some particular outcome, and can thus lead to false conclusions about the results of an experiment. This is often the type of fallacy that results in spurious claims about how everything in existence is the leading cause of cancer, along with most other ridiculous claims that are based on a sequence of events without any particular focus on other causes.
Not naming any names, of course. That would just be silly.
Let’s take the following example. Brian, after eating bacon every morning for a week, is told during a checkup that he has a tumor. A less than elegant example, I know, but the above image is making me grumpy and I just want to go cry now.
Anyways, a post hoc error would state that because eating all that yummy bacon came before the diagnosis, it must have been the cause. It wouldn’t look at details about Brian such as his chronic smoking in earlier life, or his family history of cancer. And this is the problem.
Correlation of events does not, in any way, imply causality. The sequence in which these events occured does not demonstrate any underlying connection between them, and they may in fact be entirely inconsequential in relation, with their occurence having no deeper meaning than that they happened in the first place.
In the words of Lawrence M. Krauss, “Rare things happen all the time“.
Tales from Hell 
Back to comment again as once more you’ve managed to grab my attention with a concise, informative and witty blog while I was attempting to find anything that I could comment on, in a sea of other blogs, without wanting to curl up and comfort-eat many buns.
It is very true that correlation, while unable to show either causation or directionality, is often represented as doing just that. How often over the past few decades has coffee been said to cause cancer or heart disease one week and then been touted as a preventative measure for such conditions the following week? The below article discusses the rise, fall, rise (repeat ad-nauseum until the end of time) of coffee. While criticising the studies that claimed adverse health-effects of coffee consumption, it itself implies (but crucially doesn’t claim causation) that coffee-swilling may be beneficial and advises against jumping to conclusions from correlational studies.
http://onlinelibrary.wiley.com/doi/10.1002/mnfr.200400109/abstract
Hi I really enjoyed reading your blog. You managed to make it very entertaining and informative.
It is true that it is a common occurance for correlation and causation to get mixed up in a giant melting pot. One of the most highly publised mix ups between correlation and causation was done by Dr Andrew Wakefield and the research that was done on whether the mmr jab causes autism in children (http://news.bbc.co.uk/1/hi/8268302.stm, this is an example of the articles on the topic, there were so many I didn’t quite know where to begin). This mix up, I think, clearly shows how causation can be stated as being present, when really it is just a correlation.
I look forward to your next blog 🙂
A good blog, picking up a good topic about the differentiation between correlation and causation. In my mind, the media seems to have severe difficulty understanding the differences between the two above concepts, whereby a link therefore means one must cause the other. The simple and honest fact is that correlations are quite boring, and the media don’t want to publish a headline story saying that ‘Bread might possibly be a cause of cancer, but only one study shows this and it’s not a proven causation’. It’s not quite the catchy, newspaper-selling title they might be after. This does, however, mean that those less informed on such statistical matters take such claims as ‘bread causes cancer’ at face value, and for some reason believe such ludicrous claims. And to be honest, can we really blame them? They have no knowledge of the study or what it’s about, and they don’t really want to trawl through a research paper to find the answers, so the media headline is all they have to go from.
From here then, I could go on a rant about the misrepresentations of statistics and life in the media, but I feel I have made my point enough. In some research, it’s clear to see that a correlation doesn’t mean a causation. Take the famous shark attack and ice cream example; in the summer, ice cream sales and shark attacks increase. This does not mean that ice cream sales cause shark attacks, or that shark attacks cause ice cream sales to increase. The third-variable is that of the summer, whereby in the hotter weather, more ice creams are purchased, and also more people go for a swim in the sea, increasing the chances of a shark attack. Other examples, such as those researched today, have boundaries between variables that are less clear, and to be honest, variable A could cause variable B, but the distinction is so hard to define, it’s best not to claim causation, but only correlation.
Pingback: Comments for 10th February | psud6e