Chapter 5: Experiments, Good and Bad

1
Chapter 5 Experiments, Good and Bad

• Three studies on pp. 71-72.
• Observational studies are passive data
  collections.
• Experiments are active data production.
• If properly designed, we can observe whether
  cause and effect relationships are present.

2
Definitions (p. 72)

• Response or Outcome Variable A variable whose
  changes we wish to study
• Explanatory or Predictor Variable A variable
  that explains or causes changes in the response
  variable.
• Subjects Individuals in an experiment.
• Treatment Any specific experimental condition
  applied to the units.
• For multiple explanatory variables, a treatment
  is a combination of specific values of each of
  these variables.

3
A Very Simple Experiment

• Alcohol awareness fair at the Russell House
• At one booth, students (21 years old or older)
  are given four beers (12 oz.) to drink.
• At the start and after each beer, each student is
  asked to write a given sentence on a blank piece
  of paper.
• Neatness is judged.

4
Coordination and Alcohol Experiment

• Who are the subjects?
• What is the response variable?
• What is the explanatory variable?
• What is the treatment(s)?
• Simple Experiment
• Apply a Treatment ? Look for a Response
• Problem Lacks control for many influences other
  than the treatment.

5
Lurking Variables and Confounding

• Lurking Variable A variable that is not of
  interest in a study that nonetheless influences
  the response variable. (73)
• Name some lurking variables in the previous
  alcohol/coordination example.
• Confounded The effects of two variables
  (explanatory variables or lurking variables) on a
  response variable are said to be confounded when
  they cannot be distinguished from one another.
  (73)

6
Study Web vs. Traditional Learning(pp. 73-74)
7
Randomized Comparative Experiment

• Randomized Comparative Experimental Design
  Compare the effects of a treatment on an
  experimental group to a control group. Subjects
  are randomly assigned to two groups.
• Experimental Group The collection of
  individuals (or subjects) on which the complete
  experiment is performed.
• Control Group The collection of individuals
  that is subjected to the same conditions to which
  the experimental group is subjected except that
  no treatment is imposed.
• This design helps control for some lurking
  variables and avoid problems with confounding.

8
(No Transcript)
9
Example 3 Sickle Cell Anemia (p. 75)
10
Placebo and the Placebo Effect

• Placebo A dummy medication or false treatment
  that is used to control for psychological
  effects, especially in medical experiments . (74)
• Placebo effect The response of patients to a
  placebo.
• Some people report improvement when taking
  placebo!

11
Strength of the Placebo Effect

• Angina Pain due to inadequate blood supply to
  the heart.
• Treatment in 1950s Mammary artery litagation.
  Surgery performed to tie off mammary arteries.
• Compared with placebo surgery Incision made, no
  ties made.
• Result Both groups reported pain relief at about
  same rate.

12
More than Two TreatmentsExample 4 Conserving
Energy(pp. 76-77)
13
Principles of Experimental Design (78)

• Control the effects of lurking variables on the
  response. Use a comparative design.
• Randomize - use chance to assign subjects to
  treatments.
• Replicate - use enough subjects in each group to
  reduce chance variation in the results.
• Statistical significance An observed effect so
  large that it would rarely occur by chance. (79)

14
How to Live with Observational Studies

• Good studies are comparative
• Compare two existing groups on a response
• Can use matching to control for lurking variables
• Two groups are similar except for explanatory
  variable
• Measure and adjust for confounding variables
• Example 5 (p. 80) Living longer through religion.