Can correlation show causality?

So, Blog, it’s been a while and I have to say, I did not miss you. I would much rather be lazying around watching Zorro (sexy Spanish dude) preferably with a hot water bottle. Enough about me. Let’s talk about correlation and causation. See, if it was Zorro talking about this I think my interest would certainly be engaged.

There are many definitions of correlation available, this one will suffice : “a statistical measurement of the relationship between two variables”. Correlation values range form -1 to +1 (the more negative the value, the more negative the correlation). Causation, however, is the relationship between cause and effect. Now that the definitions are out of the way, can correlation show causality?

NO.

The media likes to portray correlation studies as causation, which is wrong as people will often listen to these “proven causes” without even questioning them or looking into the research behind it. An example is a conversation I had with a friend about the implications of medicinal marijuana. She thought it was ridiculous as it has been “proved that cannabis cause schizophrenia”. I asked her… “where did you get this evidence from?” She replied “newspaper”. There is no evidence of causation, only correlation, between cannabis use and schizophrenia. This is because, there could be many other possible variables that can confound the experiment. We can never say for fact that it is one variable alone that causes any change in another. Although many statistical analysis exist, such as multiple regression, which can increase confidence when analysing the relationship, correlation still cannot prove causation.

To show causation one variable must directly cause the change in the other variable. For example, to assert causation participants must be assigned to two groups, one to measure the changed variable, and a control group. The study must have no confounding variables and high validity to suggest causation. This does not mean that correlation studies aren’t important however, as they can build the foundations of further and important studies. It also does not mean that if evidence for causation is strong it still does not mean that a change in A will definitely cause a change in B. As results from a study cannot be 100% generalizable to everybody.

http://www.experiment-resources.com/correlation-and-causation.html

Advertisements

15 thoughts on “Can correlation show causality?

  1. lmr92 says:

    I agree that we should never look at a correlational relationship and take it as proof that one variable has caused an effect in another variable. However, I don’t think this means we should totally disregard the information we get from correlations. For example, you mentioned that the relationship between cannabis and schizophrenia is only correlational, and true we can not say with certainty that cannabis causes schizophrenia, however this relationship would still put me off the use of medicinal cannabis. And if there was a history of schizophrenia in my family i would be even more swayed by this correlation.
    To sum up my point, I don’t think correlational data should be taken as fact, but neither do i think that it should be considered useless.

  2. lrowlands1 says:

    I agree that correlation, and the one in mention about schizophrenia is useful and in no way should be ignored. I was not suggesting that because it’s a correlation and not causation that it’s okay for everyone to smoke ganj. However, I do think that the media, not research, portrays this in a exaggerated way. Could I just ask out of interested, would you still be put off by the use of medicinal marijuana if you were to contract HIV and strong evidence suggests that the use of marijuana could prevent it turning into AIDS? peace x

    • psych31 says:

      yes i totally agree with you leanne, the media costantly exaggrate the importance and implications of causality when it comes to classified drugs such as cannabis and coke ; despite this they do not emphasis the harmfull chemicals which ae passed through smoking as smoking is more socially acceptable yet thesechemicalls can be more addictive ….. it is a fact that nicotine and tabacco are far more addictive than weedd but no media correlation would ever highlight this fact. xoxo

  3. tinastakeon says:

    oof, Zorro, yes please, just the thought of Antonio Banderas made me engage with your blog.

    Causation eh, well, as we know, nobody truly knows what causes schizophrenia, but I have seen the use of cannabis exacerbate symptoms of schizophrenia in someone close to me, mind you, coffee, busy buses, the sound of electricity pylons, even the bubbling of a waterfall also had strange effects on him too. So I will have to agree with you on that one, and it’s awful that the media reports such stuff. I saw a man wet himself on the bus today, he’d been drinking, and most people would assume that the alcohol was the cause of his accident. Sadly, the man on the bus has special needs, in fact he used to live with my friends parents, they were his carers as he was deemed not fit to live independently, in the last couple of years he has been living in a bungalow in the next village. I felt awfully sorry for him as the other passengers on the bus giggled at him while he stood uncomfortably waiting for the bus to stop, and made a point of telling my son why it was not funny in the hope that the other passengers would hear and be a little less judgmental. I think it’s natural for humans to make assumptions in life all the time, we just don’t have the time or capability to find out all the variables that paint the true picture of a situation. In the same way as the alcohol was not the real cause of the bus mans accident, nobody could categorically prove that cannabis causes schizophrenia. I have to agree with you that cannabis has been shown to have many and varied health benefits, but we must be careful not to make assumptions here either. Many people who take cannabis take it to get high, they are more interested in the THC content (THC is the psychoactive element of cannabis) but raising the levels of THC is done (through splicing and interbreeding and creating new strains) is done at the expense of CBD and also potentially at the expense of the more subtle healing qualities of cannabis in it’s more innocent form.

    I found this report on ways of screening samples of cannabis, and you can quite clearly see in the results section that each sample has differing levels of THC.

    http://www.estcal.com/tech_papers/papers/Security/AppnoteMarijuanna-zNose.pdf

    I feel it is important to stress that the use of marijuana for medicinal purposes ought to be under the watchful eye of a health care professional, because as you can see, every street sample is different, and the psychoactive THC could potentially intensify other problems which were previously unknown about.

    Love and peas

  4. psyalo says:

    I agree with what you have said in your blog, correlation does not always mean causation. The media do like to make us think that it does, but are often wrong, or leave things out. One site I found gave this example:
    “Children where the mother stays at home do better at school (and later in life).” http://www.cambridge2000.com/memos/correlation.html. The conclusion they are trying to give is that mothers should stay at home, to look after the child, and make them feel guilty for working. However, what they fail to mention are other variables, such as fathers.
    Another example (http://www.slideshare.net/carmean/correlation-vs-causation) mentions the belief that stress caused ulcers, however resent research found that it was actually caused by a bacteria called helicobacterpylori (H.pylori). Although stress may not of helped the ulcers, it was never the cause.
    Wright (1921), said that ”The ideal method of science is the study of the direct influence of one condition on another in experiments in which all other possible causes of variation are eliminated”.

  5. elrucron says:

    Like tinastakeon, I too have had personal experience with this. A previously healthy relative of mine developed schizophrenia as a result of marijuana use. Furthermore, at 28, they now have the mental age of someone under half their age. Would I suggest this is due to a direct result of smoking weed? Absolutely not. Would I suggest it played a part? That would be hard to deny…

    Even though I completely agree that causality is not determined by correlation, we should not ignore a correlation even if we KNOW it is not an exclusive cause of effect. Why? Because other variables may be involved that do not necessarily conflict with the original study’s dependent variable. Take weed and it’s relationship with schizophrenia. We can choose to view this as an indicator of partial causation. Other factors, such as genetic predisposition, may help to explain why some people who smoke marijuana develop mental health problems and others do not. One study found that smoking induced schizophrenia in individuals already predisposed to it *.This does not discount the original correlation, as combined we now can see a risk factor for schizophrenia, and the varying degrees of likelihood for schizophrenic development between high and low risk groups.

    *http://www.sciencedirect.com/science/article/pii/0022395694900108

  6. captkebab says:

    Whilst it’s true that causation cannot be inferred through the use of correlational designs, they frequently lead to experimental work that does. My example leads on from the previously mentioned topic of schizophrenia and cannabis use. Correlational research does suggest a relationship between drugs and self induced psychosis. In fact relationships have been found between cocaine use and even medical drugs such as L dopa with increased incidence of schizophrenia. It was through correlational research such as those mentioned that lead to the dopamine hypothesis in the eitiology of schizophrenia. This is basically that the unusual behaviour and symptoms of schizophrenia can largely be attributed to abnormalities of dopamine function in the brain.
    In turn, these and other studies paved the way for the study by Wong et al (1986) http://www.sciencemag.org/content/234/4783/1558.short.
    PET scans were used to scan for the density of D2 dopamine receptors in the caudate nucleus of healthy subjects and two groups of schizophrenic patients, one drug naive and one drug treated. It was found that the densities of D2 receptors were higher in both the drug treated and drug naive patients compared to the healthy volunteers.

  7. psud1a says:

    You’re definitely right in saying that correlation can not be used as an indicator of causation. Causation means there is a specific circumstance or stimuli causing a specific reaction to another. Correlation could infer that there are other factors and reasons in why there is that specific outcome. Even if we were to say that correlation is an indicator, how strong would the correlation have to be? Admittedly if there was a perfect positive or negative correlation score, then it could be assumed there is causation present, but further research and testing would have to be conducted.

  8. psuc98 says:

    While correlation cannot show cause and effect it is still important to research and should not be completely disregarded, in your example of cannibas causing schizophrenia you could argue that this correlational relationship highlights a future area of research, without correlation this may never even be noticed. Also as it is still unclear what causes schizophrenia people may want to avoid any possible triggers or causes, correlations are probably believed because of lack of real research.

    The real issue is when correlations are presented as a fact; for example the friend you mentioned believing this to be true could mean she/he later may turn down medicinal use of marijuana which could lead to them becoming more ill, surely it is unethical to mislead people in a way which could cause them harm. The media presentation of correlations is definitely one of the reasons people believe unsupported claims and can lead to negative consequences.

  9. raw2392 says:

    Your blog definitely got my attention by starting with Zorro! Id have to agree with psud1a and yourself by saying that correlation can not be an indicator of causation. Correlation shows that there is a relationship between the two variables but how the relationship is formed or even what affect they have on each other, simply that some how there is an effect.
    I also believe that there is a strong correlation between the used of cannabis and the onset of schizophrenia, but also believe that there are other factors which result to the onset of schizophrenia in the first place anyway, simply that the cannabis could induce the onset of the disease quicker. Take the adoption studies into schizophrenia which show that adopted children who were reared away from their schizophrenic biological parents, still got the disease, which shows that there has to be some form of genetic influence into the onset. This supports the fact that the use of cannabis cannot be fully blamed for the onset of schizophrenia because there are other factors that are linked, furthermore supporting the fact that correlation cannot be an indicator of causation.
    Great blog 🙂

Leave a Reply

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

WordPress.com Logo

You are commenting using your WordPress.com 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 )

Google+ photo

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

Connecting to %s

%d bloggers like this: