Section 3.2:
Experiments in the Real World
Equal Treatment for All in Experiments
 The experimenter must know exactly what
treatments and responses he wants information
 The experimenter must provide all materials needed
for the treatments and to measure the responses.
 All subjects must be treated alike, except for the
treatment assigned to them; any unequal treatment
can cause …
It’s not an easy task to treat them alike.
Double-Blind Experiments
 Neither the subjects nor the people who work with
them know which treatment each subject is
 Why is this so important???
If a subject knew they were receiving the placebo, the placebo
effect would be weakened and bias in favor of the other
treatment(s) would be increased.
If a subject knew they were receiving the treatment, they might
expect results that may or may not occur.
Doctors’ expectations change with how they interact with
patients and even the way they diagnose a patient’s condition
if they know which treatment a subject has received.
Double-Blind Experiments
 Until the end of the study and the results are in, only
the study’s statistician knows for sure who has the
treatment and who has the placebo.
 The cartoon brings us back
to non-sampling errors from
Chapter 2…the knowledge of
vocabulary is very
Problems that Happen During Experiments
 Sample surveys – non-response due to failure to contact
or refusal to participate
 Experiments – refusal to participate, non-adherer to
treatment, dropout
Refusal: Potential subjects may refuse to participate, which can
cause undercoverage of certain groups in the subject population.
Non-adherers: Subjects participate, but don’t follow the
experimental treatment as defined. They may add or omit other
medications which may or may not be part of the experiment.
Dropouts: Subjects leave the experiment before its conclusion. If
reason is not related to the treatment (they move away), no bias has
occurred. If the reason for dropout is related to the treatment
(reaction to treatment), bias can result.
What do you think???
What will Agent B be doing
What will Agent Q be doing
What will Agent K be doing
B = treated differently
Q = non-adherer
K = dropout
Your Homework…
 Find print (magazine/newspaper) or online info regarding a clinical
 Print or photocopy entire article and bring to class on Friday. Make
sure it has information about the procedures of the experiment.
 Again, this is due in class on Friday.
 Tomorrow, we will be passing back papers and passing out the
books which we have. Those which do not get a book will get a
photocopy of Chapter 3. I have been told books will be here this
week (cross your fingers).
 If you have borrowed a book, with permission or not, please bring it
tomorrow so I can properly check it out.
 Quiz for 3.1 is on block day.
HW Turn In #5
 #3.1A (state yes or no and give reason)
 #3.6A (state the response variable)
 #3.9B (state the first 5 rooms to the flat-rate group)
 #3.14 (state 1 lurking variable)
Section 3.2:
Experiments in the Real World
Day 2
Completely Randomized Experimental Design
 All subjects are randomly assigned to groups, and all
groups are given different treatments.
 So far the examples have had only one explanatory
variable (ex., drug vs. placebo). A completely
randomized design can have any number of
explanatory variables…
Example of Two Explanatory Variables
In this example, there are two explanatory variables to describe the durability
of fabric under repeated washings. The type of cleansing agent and the
temperature of the water are both explanatory variables which are being
tested. This produces 9 different treatments.
A combination effect is called an interaction.
Matched Pair Designs
 Combines matching with randomization. It is an
example of block designs.
 Compares just two treatments.
 Choose pairs of subjects as closely matched as
possible. Assign one treatment to each subject by
tossing a coin or reading odd and even digits from
Table A. (every heads goes to group 1, every tails to
group 2).
 Sometimes a matched pair is one person who tests
two items, one after the other.
Block Design
 A block is a group of experimental subjects that have
some commonality that is known before the
experiment that could affect the response to the
treatments (could be divided by gender, age range,
 In a block design, the random assignment of subjects
to treatments is carried out separately within each
block (like two, or more, randomized comparative
Example of a Block Design
What type of sample design is a block design similar to?
In Your Groups
 In your groups, complete the following problem:
Most motor vehicles are equipped with catalytic converters to reduce
harmful emissions. The ceramic used to make the converters must
be baked to a certain hardness. The manufacturer must decide
which of three temperatures (500°, 750°, and 1000°F) is best. The
position of the converter in the oven (front, middle, or back) also
affects the hardness. So there are two experimental factors:
temperature and placement.
A. List the treatments in this experiment if all combinations of levels
of the two factors are used.
 B. Design a completely randomized experiment with five units in each
 C. Using Table A, beginning at line 101, do the randomization for your
Homework Assignment #11
159 #3.24, 3.25
Pages 163-164, #3.27, 3.29, 3.30
Pages 168-169, #3.32, 3.33, 3.34
to bring your clinical
study article to class on Friday.

Section 3.2: Experiments in the Real World