Repeat after me: “Correlation does not imply causation”

The post hoc ergo propter hoc fallacy is one of the great weapons in the arsenal of denialists. The reason it works so well is it makes sense. As my readers know, my daughter is dealing with a nasty respiratory virus. One of the doctors told my wife, who is not a medical professional, that kids with this virus go on to develop asthma. My wife was not pleased to hear this. What the doc didn’t tell my wife was whether or not there is a causal relationship between the infection and asthma. It is also true that kids who get RSV end up going to school.

Asthma is a common illness. RSV is more common. Some prospective studies have followed children who have had RSV and looked at the rates of asthma, and found them to be rather high. The problem is, most children get RSV. It is very difficult to prove that a ubiquitous virus has a causal role in a common respiratory disease. The best study would follow similar children with and without RSV, but where do you find kids without RSV? How do you know that kids who don’t have clinically apparent RSV infections aren’t just resistant to respiratory disease in general? That being said, there is a fascinating literature on the effect of RSV on the lungs and the immune system. I’m actually becoming a living laboratory as we speak.

Human beings can’t help being susceptible to the causation/correlation fallacy. It’s probably hard-wired. But we can watch out for it and maintain a skeptical eye on claims of causation.


Comments

7 responses to “Repeat after me: “Correlation does not imply causation””

  1. It’s also important to remember that even though there may not be a causal relationship, correlation does imply a relationship. Sleeping with your shoes on may not be the cause of your headache the next morning, but they’re related by the fact that you had too much to drink the night before.

    I notice that global warming deniers, when faced with the correlation between CO2 and global temperature, with a correlation coefficient of 0.93 between CO2 and 5-year average temperature (I have the code if anyone wants it), use the “correlation does not imply causation” canard to try and dismiss any relationship.

    Correlation, when it is used to test an already scientifically sound hypothesis of causation, may not “prove” anything, but it can be used as evidence that the hypothesis is correct. I really think that can be put better, but I think you see what I mean.

  2. I see you experienced the very kind useless advice from the “little knowledge is dangerous group”. They seem to come out of the woodwork once you have a child with any kind of issue.

    In case you’ve not seen me much on Orac’s blog: I am a mom to a kid who had neonatal seizures, who had a later seizure during a nasty gastrointestinal illness with a trip by ambulance to the hospital, then had four separate stays in the hospital for croup (one time during Thanksgiving), and who has a severe speech disability, and now as a young adult gets migraines and has hypertrophic cardiomyopathy.

    Whew!

    Useless information includes:
    Don’t drink milk
    Don’t eat wheat (apparently according to one guy they both cause seizures)
    Don’t vaccinate (someone tried to blame his newborn seizures on the HepB, he got that when he was 6 years old!… that was because his baby sister was the first of our family to get that vaccine)
    Don’t let him sleep on his back (uh, before the SIDS recommendation, it turns out as a baby he hated being on his tummy and could push himself from tummy to back at age 2 months)
    You don’t talk to him! (oh, yeah, right… I actually played foreign language tapes to him hoping he would be a language genius, little did I know he would need ten years of speech/language therapy to communicate in English!)
    Don’t use sign language (um, that was a break through! It really helped find out there was a real boy in that wordless body).

    and on and on and on and on.

    You can see why the Mercury Militia is made up of parents who are looking to blame, or are susceptible to such “advice”. They are looking for answers, even when there are no real answers.

    Even my son’s child neurologist refused to pin his speech disability on the seizures. He would say that they “may or may not” be a reason for the “static encephalopathy” (which is neuro-speak for “we know something is wrong, but not what, and it is not changing”).

    But one note: the child neurologist has a child with speech problems, which required speech and language therapy… so despite being kind of odd in a super intellectual way (all but one neurologist I have dealt with could be categorized as medical uber-geeks), he understood what it was like to have a child with speech/learning issues.

    I wish the best to you and your family. From personal experience, I know that your experience was stressful, and I’m sure your daughter’s hospitalization will affect how you treat your patients in a very positive way. I am so glad she recovered.

    Though, I am sorry to say: She will turn into a teenager.

    I have a soon to be 14 year old daughter. She has been more trouble than both boys put together! One reason is she is so bright, I blame that on her getting the HepB vaccine at birth.

    One the good side: we both share the same strange sense of humor. Joy comes from walking down the street getting stares because of our strange and bizarre conversation.

  3. It is ironic that you should blog about the “post hoc/cum hoc, ergo propter hoc” fallacies just as I am combating similar reasoning in two different fields!

    a) In Software: just cause you have two faults doesnt imply that one was the cause of the other etc.

    b) A recent study titled “You are what your mother eats: evidence for maternal preconception diet influencing foetal sex in humans” (doi:10.1098/rspb.2008.0105) which suggests that eating an extra fruit pastille a day (by calorific content) pre conception can increase your chances of having a baby boy significantly.

    As you rightly say – to seek a causal inference is somehow hardwired in our psyche’s. I just wish people would look for evidence to draw conclusions rather than look for evidence to support conclusions.

  4. Those are wise words and are indeed describe a two way street.

  5. jnovitz

    ” notice that global warming deniers, when faced with the correlation between CO2 and global temperature, with a correlation coefficient of 0.93 between CO2 and 5-year average temperature (I have the code if anyone wants it), use the “correlation does not imply causation” canard to try and dismiss any relationship. ”

    Its fairly easy to demonstrate with the historic data that the correlation between CO2 and global temperature can be explained by the fact that rises in global temperature causes increases in atmospheric CO2 rather than vice versa.

    That does not mean that the opposite couldnt happen, simply that the historical data up til the industrial age is best explained by temperature rises causing CO2 rises.

    “I’m actually becoming a living laboratory as we speak.

    Good Lord!

    “The best study would follow similar children with and without RSV, but where do you find kids without RSV? ”

    Actually its quite simple. You treat kids with a monoclonal antibody against RSV (about 80% effective) and compare rates of development of recurrent wheezing versus a control group.

    In anycase the claim was that a severe reaction to RSV (ie requirement for hospitalisation) might associate with a susceptibility to asthma rather than just an ordinary infection.

  6. jnovitz, for your information, the correlation coefficient between fossil fuel consumption and the increase in atmospheric CO2 over the last 120 years or so is more than 0.99. I didn’t mention that to keep my comment short and to the point, thinking that surely nobody was stupid enough to suggest that global warming is increasing CO2. My mistake. But don’t take my word for it. The data is freely available. Analyze it yourself. Pasted directly from the program I wrote to get the correlation:

    % The temperature data is obtained from the Goddard
    % Institute for Space Studies at NASA
    % http://data.giss.nasa.gov/gistemp/

    % The column of atmospheric carbon concentration has two
    % sources. Data from 1880 to 1958 comes from the smoothed
    % Law Dome DSS antarctic ice core sample data available at
    % http://cdiac.esd.ornl.gov/ftp/trends/co2/lawdome.combined.dat
    %
    % From 1959 to 1999 comes from the Mauna Loa Atmospheric
    % Carbon Dioxide data available at
    % http://cdiac.esd.ornl.gov/ftp/maunaloa-co2/maunaloa.co2

    % The fifth column is annual CO2 emissions from fossil
    % fuels. The sixth are cumulative emissions (since 1751).
    % Both are in millions of metric tons. Data are from
    % http://cdiac.ornl.gov/trends/emis/tre_glob.htm

  7. jnovitz

    “I didn’t mention that to keep my comment short and to the point, thinking that surely nobody was stupid enough to suggest that global warming is increasing CO2.”

    The fact that over the last 400 000 years global warming (ie increases in solar radiation received due to eccentricities of orbit) has correlated with increasing CO2.

    Now if you wish to argue that increasing CO2 causes eccentricities in the earth’s orbit around the sun….be my guest.

    http://en.wikipedia.org/wiki/Image:Vostok_420ky_4curves_insolation.jpg

    I will however prefer to explain it by warming oceans releasing CO2 due to standard thermodynamics.

    This does not NECESSARILY mean that the data over the last 100 years is due to same cause.

    But there can be little doubt that the correlation over the last 400 000 years is due to increases in global temperatures driving increases in CO2, not vice a versa.

Leave a Reply

Your email address will not be published. Required fields are marked *