Page 281 - Statistics for Dummies

Chapter 17: Experiments: Medical Breakthroughs or Misleading Results?
Designing the experiment
to make comparisons
Every experiment has to make bonafide comparisons to be credible. This
seems to go without saying, but researchers often are so gung-ho to prove
their results that they forget (or just don’t bother) to show that their factor,
and not some other factor(s), including random chance, was the actual cause
for any differences found in the response.
For example, suppose a researcher is convinced that taking vitamin C pre-
vents colds, and she assigns subjects to take one vitamin C pill per day and
follows them for 6 months. Suppose the subjects get very few colds during
that time. Can she attribute these results to the vitamin C and nothing else?
No; there’s no way of knowing whether the subjects would have been just as
healthy without the vitamin C, due to some other factor(s), or just by chance.
There’s nothing to compare the results to.
To tease out the real effect (if any) that your factor has on the response, you 265
need a baseline to compare the results to. This baseline is called the control.
Different methods exist for creating a control in an experiment; depending on
the situation, one method typically rises to the top as being the most appro-
priate. Three common methods for including control are to administer: 1) a
fake treatment; 2) a standard treatment; or 3) no treatment. The following sec-
tions describe each method.
When examining the results of an experiment, make sure the researchers estab-
lished a baseline by creating a control group. Without a control group, you
have nothing to compare the results to, and you never know whether the treat-
ment being applied was the real cause of any differences found in the response.
Fake treatments — the placebo effect
A fake treatment (also called a placebo) is not distinguishable from a “real”
treatment by the subject. For example, when drugs are administered, a
subject assigned to the placebo will receive a fake pill that looks and tastes
exactly like a real pill; it’s just filled with an inert substance like sugar
instead of the actual drug. A placebo establishes a baseline measure for
what responses would have taken place anyway, in lieu of any treatment
(this would have helped the vitamin C study mentioned under “Designing
the experiment to make comparisons”). But a fake treatment also takes into
account what researchers call the placebo effect, a response that people have
(or think they’re having) because they know they’re getting some type of
“treatment” (even if that treatment is a fake treatment, such as sugar pills).
Pharmaceutical companies are required to account for the placebo effect
when examining both the positive and negative effects of a drug. When you
see an ad for a drug in a magazine, you see the positive results of the drug

3/25/11 8:13 PM
25_9780470911082-ch17.indd 265 3/25/11 8:13 PM
25_9780470911082-ch17.indd 265