iii. Influence of extraneous or confounding variables
A confounding variable is another variable whose effect
on the response variable cannot be separated from the explanatory variable
under study.
To examine the effect that confounding variables can have on data, examine
the table below that comes from the same study into anaesthetics reported
in the previous scenario.
Confounding variables in clinical trials
The placebo effect
Confounding variables can be very important in clinical trials where
a new drug or procedure is being tested. Let’s say a doctor gives
a patient a new drug in tablet form and the patient gets better. How can
you tell whether it was
1) the attention that was given to the patient as the drug was administered,
or
2) the drug itself that caused the improvement? Many patients respond
positively to any treatment, even when they are given a placebo,
i.e. a dummy medication. In other words, it is the process of being treated,
not the action of the drug, which produces patient improvement. As a result,
it becomes important to separate the drug (explanatory variable) from
the treatment (confounding variable). An improvement in a person's health
that occurs when they are given a dummy medication is called the
placebo effect.
Observer bias
Another effect that can confound the results of an experiment is related
to the expectations of the tester. If the doctor administering the treatment
knows which patients are being given the drug and which are being given
the placebo, the doctor might note improvements for those patients on
the drug and no improvement for those on the placebo. This is called observer
bias.
Identify the explanatory variable and the most likely confounding variable
in the following scenarios.
SCENARIO
Last year 20% of a group of adult women did not have a cold throughout the year. This year they all took Echinacea capsules every day and
30% did not get a cold. One of the participants in the study claimed that these results showed that Echinacea prevents colds.
SCENARIO
A study of engineers showed that those who had completed a certificate earned 10% more, on average, than those who had completed a degree.
