This page is intended to help solidify the following enduring understandings:
From the WNYC Radiolab website:
With new research demonstrating the startling power of the placebo effect, this hour of Radiolab examines the chemical consequences of belief and imagination. Could the best medicine be no medicine at all? We take stock of the pharmacy in our brains, consider the symbolic power of the doctor coat, and visit the tent of a self-proclaimed faith healer.
What is an experiment?
One of the more common mistakes that people often make with regard to interpreting data is confusing correlational studies and experimental studies. Correlation is the degree to which a change in one variable generally associates to a change in another variable. Causation is the direct impact of one variable on another.
For example, let's look at a couple of news items related to a study on the relationship between television viewing and life expectancy.
Every hour of TV watching shortens life by 22 minutes
Time watching television may shorten life, study says
Based on these articles, I might be led to believe that if I were to sit down to watch a Giants game for three hours, I would shave an hour off of my life expectancy. That is a mischaracterization of the study's results, however. While the researchers make a case that there is an association between the two, they stop short of saying that one causes the other.
Here's the original study. Take a look at the abstract and pay particular attention to the last sentence in the results portion: "This study is limited by the low precision with which the relationship between TV viewing time and mortality is currently known." One of the reasons the researchers can't say for sure that tv watching causes shortened life expectancy is that there may be (and surely are) lurking variables that might come into play. For instance,
Does watching educational programming help prepare students for greater success in formal education, which according to this study might increase life expectancy?
Does watching violent programming lead to unhealthy behaviors as this researcher claims?
The fact is, we don't know the relationships between all of these variables. We can use what we know from these studies and others to form some pretty good hypotheses that we might then test, but we wouldn't know for sure unless we were to conduct an experiment. In the meantime, we are wise to remember this mantra:
Correlation does not imply causation.
Examining causation is the work of experiments, which we examine further below.
Experimental design is the method by which we determine the effect of the manipulation of one variable on an outcome. Experimental design is the only research design where we can establish causation because we work to ensure to the greatest degree possible that conditions for the experimental group and the control group are as close to the same as possible, with the exception of one variable.
The experimental group is the group that is exposed to the experimental condition. That is, this group is exposed to the condition we wish to test.
For example, if we are testing whether classical music helps students retain more information while studying, the experimental group would be the group that listens to classical music.
The control group is the group that is exposed to the baseline condition and does not receive the experimental intervention. They may, however, receive an inert intervention, such as a placebo, so that they are unaware of what group they are in.