Observational science versus Experimental science

We often hear the claim that the debate is over, that human activity (CO2) is responsible for changing the climate, that there is overwhelming evidence to prove this. But can this level of certainty in attribution really be achieved in climate science? As far as I know climate science is an “observational” science. This kind of science is being done through the observation of nature just taking its course and recording the findings over time. As opposed to the “experimental” sciences that advance via controlled experiments.

There are other observational sciences like geology, social sciences, epidemiology and so on. All those sciences can not perform controlled experiments because it would not be ethical or it would be impossible to do. For example, in astronomy it would not be possible to manipulate stars or galaxies in order to observe what happens when the researcher changes something. Or in medical observational science it would not be ethical to administer one group with a medicine, while another group gets only a placebo.

In climate science there is only one test subject: Earth. We don’t have the luxury of three similar planets in which we extract CO2 from one, add CO2 to the other and compare that to our Earth. Therefor there is no other option than to observe what is happening and record the findings.

Observational sciences have their limitations. Observing nature taking its course means there are many variables involved and in the research only one (or a few) are accounted for. To pinpoint one cause, and only one, it is necessary to isolate that factor and change that while keeping other factors the same. Then if there is a change in output, one could be sure that this is because of this one factor that was changed. That is what experimental science does and that is the difference with observational science.

The consequences are in the extent to draw conclusions about causality. There is always the risk that some variable that wasn’t observed is the “real” cause, therefor wrongly assuming that a measured variable is the cause. While in reality the cause is one of the unmeasured variables. This is not well understood and therefor results of observational studies are easily whipped up in the media.

A classic example of the difference between observational and experimental studies is the use of the prescription of estrogen to woman countering menopause and the chronic ills of aging. An observational study was done in the mid-1980s and followed a large group of nurses. The conclusion of the study was that those who took the estrogen had only one third as many heart attacks than those who never took the drug. By the mid-1990s the beneficial effects on heart attacks by administering estrogen were thought to be well established.

A couple years later an experimental trial called HERS (Heart and Estrogen-progestin Replacement Study) found that estrogen therapy INcreased, in stead of decreased, the risk of a heart attack. This was confirmed by a later clinical trial.

Why did the observational study came to the conclusion that estrogen was decreasing the risk of an heart attack, while the experimental trials came to the opposite conclusion? It had to do with the the subjects. In the observational study the subjects were nurses who were traditionally more health conscientious, fitter, have fewer risk factors for heart disease, were wealthier and more educated than a lot of other groups of women.

The effect found in the observational study had its origin in the test group which had a reduced risk for heart disease in the first place and could handle the negative effects of the estrogen better. Therefor it gave an unwarranted positive result.

In the observational tests the investigators thought they found the effects of estrogen, while the real cause was something else they didn’t look at (a test group that had a lower risk than the general population to begin with). Therefor they came to the wrong conclusion.

This shows the limitation of an observational study: without knowing the full and complete background of the subjects, there is no way of determining whether other factors were involved.

In the experimental study all those variables were kept as constant as possible and only the tested factor was changed. Therefor it was possible to find the real effect of the estrogen, while this was not possible in the observational study.

Now back to our CO2 story. Our climate researchers in the media are telling us that they could pinpoint the real cause of global warming and they are really sure about it. The system they are investigating is very complex with many variables that all could be (partly) responsible. Because they only have one subject, they only can do observational trials, in which it is dangerous to assume cause and effect relationships. Yet they are so very sure.

Just as the experts at the end of last century were also so very sure that the beneficial effects of estrogen were established…


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 )

Connecting to %s