Designing Experiments
Gathering Useful Data For Examining Relationships
Unit, single individual or object being measured
Experimental Unit, basic entity (person, plant, etc.) to which
the experiments can be assigned to
When experimental units are human beings, this is known as a
Subject. Also, known as participants
Factors, explanatory variables, “ in general this is what the
researcher is studying.”
Treatment/Level, specific experimental condition or value
for the factors
Designing Experiments
Gathering Useful Data For Examining Relationships
Randomized Experiments in more detail…
Researchers often recruit volunteers to participate in randomized experiments. These volunteers agree to recieve whichever treatment is assigned to them , and
they often don’t know which treatment it is until the experiment is concluded.
Purpose of a random assessment is to make the groups approximately equal in all respects except for the explanatory variable, which is strategically
manipulated
Pros of an experiment over an observational study…
Experiments give good evidence for causation (cause and effect)
Allow us to study specific factors researchers are interested in, while
controlling the effects of the lurking variables. *
Elements Of A Experimental Design
Placebo, “a dummy pill”-has no active ingredient or
effect
Placebo Effect, when participant/subject believes that he
or she have been “cured” simply because they think it should happen not because they were given anything
Find the subject, treatments, and response variable…
I. The ability to grow in the shade may help pines trees resist drought. How well do pines grow in the shade? Investigators planted pine seedlings in a greenhouse in either full light or light reduced by 5%. At the end of they study, they dried the young trees and weighed them.
II. A food manufacturer uses package liners that are sealed at the top by applying heated jaws. What effect does the
temperature of the jaws have on the force required to peel the liner? To answer the question, the engineers prepare 20 pairs of pieces and seals 5 pairs at 4 different temperatures then measure the strength needed to peel each seal.
Principles of Experimental Design
i. Control, in experiments the researcher attempts to manage the effects of the situation in order to minimize lurking variables
ii. Randomize, use chance to assign experimental units to treatments allows us to create a better experiment
iii. Replicate, each treatment is given to many units in order to reduce the chance variation in the results, and create the same result multiple times, therefore allowing the researcher to come to some conclusion
i. These conclusions are made through using Statistical Significance
i. An observed effect so large that it would rarely occur by chance
ii. Tells the consumer that the investigators found good evidence for the effect they were seeking
Different Experimental Designs
Randomized Comparative Design
Block Design
Matched Pairs Design
Random Allocation
Group 1 (10 students)
Treatment 1: Answer questions
Compare Correct Answers
Group 2 (10 students)
Treatment 2:
Answer questions with distractions
Chapter 5 Section 2
Ran
dom
ized
Com
para
tive
Pros And Cons Of Randomized Comparative Design
Randomization allows the two groups to be similar in all
respects
Comparative allows us to know that the treatment is the true
cause
Results are often bias
Simple design – which provides invalid data
No control group
Control Group:
patients who receive no
treatment but might think that they do
Problem to Ponder:
We are going to do an experiment in which we study the
effects of different methods of quitting smoking. We find 75 people that are willing to participate. The three
methods of quitting are going to be by pill, patch and a placebo pill. We will randomly assign 25 people to each treatment.
Draw an outline of this design. How would you use the table of random digits? Random Assignment
Definition
Block Design, random assignment of units to treatments is
carried out separately within each block
Blocks, experimental units(subjects) divided up into
homogeneous groups, and each treatment is randomly assigned to one or more units in each block.
Can have blocks of any size Male vs. Female
Smoker vs. Non-Smoker
Subjects
Males
Group 1 With Picture
Compare Results
Group 2 No Picture
Females
Group 1 With Picture
Group 2 No Picture
Compare Results
Chapter 5 Section 2
Bloc
k De
Pros And Cons Of Block Design
Combines the idea of creating equivalent treatment groups,
but treatment is random
Controls the effect of some outside variables
Able to block based on variables you cannot control
Able to draw separate conclusions
More precise overall conclusions
Remember this: Control what you can, block
what you can’t control and randomize the rest!
Matched Pairs Design
Special case of block design
Used only on experiments that have two treatments
Subjects are grouped into pairs of similar type (age, gender,
income …)
Within each pair subject is randomly assigned a different
treatment
Matc
hed P
airs
Desig
n
Chapter 5 Section 2
Pros And Cons Of Match Pairs
Compares only TWO treatments
Illustrate clearly the principles of Control, replication &
randomization
Still open for bias because of selection of treatments
Draws more precise conclusions about treatment
Draw separate conclusions about each block
Cautions About Experiments
Good experiments require careful attention to detail
Be Aware Of…
When the researcher who is measuring the reaction of the
patient were to know which group they were in, the measurements could be taken incorrectly.
When the subject who is participating in the experiment
knows which group he or she is in, the response of the effects is exaggerated, lied about and falsely taken on.
Cautions About Experiments
Thus Encouraging …blinding
Single-blinded, experiment is one in which the subjects do
not know which treatment he or she has been assigned to.
Double Blinded, neither subjects or personnel knew which
treatment was received
Lack of Realism can prevent us from generalizing the results
The more replication the more accurate the result
Chapter 5 Practice
Marketing researchers wonder if the color and type of candy’s packaging may influence sales of the candy. They manufacturer test packages for chocolate mints in three colors (white, green and silver) and three types (box, bag and roll). Suspecting the sales may depend on a combination of package colors and type, the researchers prepare nine different packages, then market them for several weeks in convenience stores in various
locations. In this experiment….
What are the experimental units?
What are the factors?
What are the treatments?
A recent study evaluated elementary age children
for aggressiveness. This study found that the
children who played video games were more likely to engage in aggressive or violent play at school. The researchers said the difference was statistically significant.
Briefly explain what “statistically
significant” means in this context.
The news media reported that this study
proved that playing computer games
causes children to be aggressive or
violent. Briefly explain why this
Name and describe the kind of bias that might be
present if administration decides that instead of subjecting people to random testing they’ll just…
Interview employees about possible drug
abuse.
Does cinnamon get our brain working faster for
males? We decided to test this idea through an experiment comparing three different treatments
(cinnamon, no cinnamon, placebo cinnamon) treating sex as the blocking variable. Two separate
randomizations are done, one where the males
receive the treatment, the other where the females receive the treatment. We have 300 males and 300 females volunteer to participate in this study.