Warm-up

A newspaper article about an opinion poll says
that “43% of Americans approve of the
president’s overall job performance.” Toward
the end of the article, you read : “The poll is
based on telephone interviews with 1210 adults
from around the United States, excluding Alaska
and Hawaii.”
– What variable did this poll measure?
– What population do you think the newspaper wants
information about?
– What was the sample?
– Are there any sources of bias in the sampling method
used?
Section 4.2
Designing Experiments
Experiment vs. Observational Study
Recall: in an experiment, we actually do
something to people, animals, or objects
in order to observe the response.
 Observational studies observe individuals
and measure outcomes but do not impose
treatment on the individuals.

Terminology

Experimental units: the individuals on
which the experiment is conducted.
– When units are people, we call them subjects
or participants.

Treatment: the specific experimental
condition applied to the units.
– Example: a treatment might be 500 mg of
naproxen, or 30 minutes of exercise per day…
Terminology
It is important to note the difference
between response and explanatory
variables in an experiment.
 Factor – another name for explanatory
variable
 Level – a specific value of a factor (ex.
dosage of medication)
 Placebo – a “dummy” treatment

Example – Physician’s Health Study
Beta Carotene



Yes
Yes
No

Does regularly taking
aspirin or beta carotene
help reduce the risk of a
heart attack?
Subjects: 21,996 male
physicians
Factors: Aspirin and Beta
Carotene
Levels for each: Yes/No
Treatments: Four options
Aspirin

No
Aspirin and Aspirin and
Beta
placebo
Carotene
Beta
Carotene
Placebo
Aspirin and
Beta
Carotene
Placebo
Aspirin and
Placebo
Beta
Carotene
Some ways to make experiments
better
You need to have a control group. This is
a group in which no treatment is given but
all other aspects of the situation are the
same.
 The placebo effect… Some patients
respond favorably even if they are not
receiving treatment but think they are.

Comparative Experiments
Example of Single Treatment
Units
Treatments
Observe/Measure Response
Single treatment designs are not optimal. There is
no control group, and the design doesn’t control for
placebo effect (when patients expect relief or
improvement).
Comparative Experiments
Using a control group
Group 1
Treatment 1
Compare Response
Units
Group 2
Treatment 2:
Placebo
The question is HOW DO WE DECIDE
WHICH UNITS RECEIVE WHICH
TREATMENT???
The answer is
RANDOMIZATION. You
MUST label this if using a
diagram and make a
statement as to how you are
going to randomize!
Principles of Experimental Design –
Things you should consider and discuss
in your answers.
Control – comparison of several treatments in to
a group without treatment is the simplest form.
 Randomization – randomly assign subjects to
treatment groups in order to reduce systematic
differences among the groups.
 Replication – Replicate each treatment on many
subjects to reduce chance variation in the
results. You are more likely to find statistical
significance if you have more people.

Illustration of Completely
Randomized Design
Group 1
Treatment 1
Random
Assignment
Compare Response
Group 2
Random
Assignment
illustrates the
principle of
RANDOMIZATION.
Choosing an
adequately large
sample ensures
REPLICATION.
Treatment 2:
Placebo
Having a control
group illustrates the
principle of
CONTROL.
Caution

There is a difference between random
selection of participants and random
assignment of subjects to treatment
groups.
– Choosing an SRS is random selection. All of
these participants are in the experiment.
– Then, randomly assign those subjects to
treatment groups.
Example 2
A pharmaceutical engineer is studying the effects of a
new medicine for pain relief. She wants to try two
different dosages (500 mg and 1000 mg) and 3 different
daily intakes (1/day; 2/day; and 3/day). Five people will
be tested at each combination of dosage and daily
intake.
 Identify all the explanatory variables.
 How many subjects are needed?
 Outline in diagram form an appropriate design for this
experiment. Indicate how many people are assigned to
each treatment group
 Use Table B starting at line 108 to select the people
assigned to the 1st treatment group. How did you label
the people?

Cautions about Experimentation
We need to treat all the experimental units in
the exact same way except for the difference in
treatment.
 One way to accomplish this is by using a
“double-blind” experiment.
– Neither the subjects nor the personnel who
have contact with them (especially the ones
collecting the data) are aware of which
treatment each subject receives.

Other Weaknesses in
Experimental Designs

Lack of realism is a serious potential
weakness.
– In 1986, before a third brake light was
required on cars, an experiment found that
adding the third brake light to cars would
reduce rear-end collisions by 50%
– In actuality, the reduction was only 5%.
What happened???
Another Favorite of the AP Gang
– Matched Pairs
Gives individuals both treatments – very
favorable because you can see differences
for each person.
 Example: You want to know what type of
music makes students perform better on
tests. You could give each student the
same test twice but with different music
each time. The order of test would be
random.

Block Designs
Grouping by another factor first.
 Example: You might “block” a cancer treatment
study into gender.
 Block designs can have blocks of any size (ex.
20 females and 50 males in the experiment).

– In a block design, the random assignment of units to
treatments is carried out separately within each block.
General Format of a Blocked
Design
Treatment 1
Block 1
Treatment 2
Observe/Compare
Response Variable
Treatment 3
Subjects
Treatment 1
Block 2
Treatment 2
Treatment 3
Putting
subjects into
blocks is NOT
random.
Observe/Compare
Response Variable
2002 #2

A manufacturer of boots plans to conduct an experiment
to compare a new method of waterproofing to the
current method. The appearance of the boots is not
changed by either method. The company recruits 100
volunteers in Seattle, where it rains frequently, to wear
the boots as they normally would for 6 months. At the
end of the 6 months, the boots will be returned to the
company to be evaluated for water damage.

A) Describe a design for this experiment that uses the
100 volunteers. Include a few sentences on how it
would be implemented.

B) Could your experiment be double-blind? Explain.
Homework
Chapter 4
# 46, 50, 54, 56, 62