Transcript Lecture 12 - Understanding Experiments

Understanding
Experiments
Lecture 12
Section 3.5
Tue, Feb 6, 2007
Example
Suppose a drug is given to 100 patients
suffering from a particular disease.
 After 2 weeks, 90% of the patients have
recovered.
 The researchers conclude that the drug
was effective.
 What is wrong with this?

Random Allocation





The assignment to the groups should be random
(a randomized design).
The subjects could be numbered 1 – 100.
Then use a random number generator to obtain
50 (distinct) random numbers from 1 – 100.
Those 50 subjects would be assigned to Group
1.
The rest would be assigned to Group 2.
Treatment and Control Groups
Treatment group.
 Control group.

Random vs. Nonrandom Allocation
Would it be wrong to allow the individuals
choose whether to be in the treatment
group or the control group?
 Would it be wrong for the researchers to
decide, subject by subject, who goes in
which group?
 Why?

Response Bias

Are the subjects in the treatment group aware of
the purpose of the experiment?
 If

Are the subjects in the control group aware that
they are not receiving the drug?
 If

so, will they be more likely to report feeling well?
so, will they be more likely to report feeling ill?
What can we do about this?
Placebos

Placebo.
Placebos
Placebos
Everybody in the treatment group gets is
administered the drug.
 Everybody in the control group gets the
placebo.
 The researchers look for differences in the
groups’ recovery rates.

Blinded Experiments
Of course, the subjects should not know
whether they are receiving the placebo or
the genuine treatment.
 Single-blind experiment

Experimenter Bias
Experimenter bias
 This is similar to response bias except that
it is caused by the experimenter, not the
subject.
 How can we control for experimenter bias?

Double-Blind Experiments

Double-blind experiment
The Placebo Effect

The placebo effect is when a subject
responds to the “treatment,” even though
the treatment was only a placebo.