EXPERIMENTS AND CAUSALITY

1
EXPERIMENTS AND CAUSALITY
2
Examples of causality questions

• The Mozart Effect

• Night lights

3
The Mozart Effect

• People who listen to classical music have a
  higher IQ than people who do not.

4
VARIABLES Independent and Dependent

• CAUSE
• Independent variable
• OUTCOME
• Dependent variable

CAUSE
OUTCOME
5
VARIABLES 3rd Variables

• CAUSE
• OUTCOME
• 3rd Variable

3rd Variable
CAUSE
OUTCOME
6
3rd Variables?

• What ELSE might be related to
• Listening to classical music
• and
• Intelligence?

7
Establishing causality

• Surveys allow you to examine covariation.
• Statistical techniques control for third
  variables
• But experiments offer control over timing
• Manipulate the Independent Variable and observe
  the consequences (the Dependent Variable).
• Select a group of subjects
• Do something to them
• Observe the effects

8
Experimental test

• Steele KM, Brown JD, Stoecker JA.
• Failure to confirm the Rauscher and Shaws
  Mozart effect.The Mozart effect is an increase
  in spatial reasoning scores detected immediately
  after listening to the first movement of a Mozart
  piano sonata (Rauscher and Shaw, 1998)
• This experiment attempted to replicate the
  unpublished study.
• 206 college students were exposed to one of three
  sequences, pretest-Verbal distractor
  material-Mozart, pretest-Mozart-Verbal distractor
  material, and pretest-Verbal distractor material.
  
• An immediate posttest indicated no significant
  difference on solution of paper folding and
  cutting items among the three groups.
• The results do not support Rauscher and Shaw
  (1998). Our negative results are consistent with
  prior failures in other laboratories to produce a
  Mozart effect.

9
Quinn, G.E., Shin, C.H., Maguire, M.G. and Stone,
R.A. (1999) Myopia and ambient lighting at night.
Nature, 399, 113-114.

• Researchers at the University of Pennsylvania
  tested the vision of 479 children between the
  ages of 2 and 16 years old. The parents of these
  children were asked the following questions about
  the lighting conditions in the rooms of their
  children when the children were younger than 2
  years old
• Did your child sleep in darkness?
• Did your child sleep with a night light on?
• Did your child sleep with a room light on?
• RESULTS
• 10 of the children who slept in darkness were
  nearsighted.
• 34 of the children who slept with a night light
• 55 of the children who slept with a room light
  were nearsighted.

10
Replications

• The study published in 1999 did not ask parents
  if they were nearsighted and therefore, this
  important piece of information was missing.
• Gwiazda, J., Ong, E., Held, R. and Thorn, F.,
  Myopia and ambient night-light lighting, Nature,
  vol. 404, page 144, 2000.
• Stone, R.A., Maguire, M.G. and Quinn, G.E.,
  Myopia and ambient night-light lighting, Nature,
  vol. 404, page 144, 2000.
• Zadnik, K., Jones, L.A., Irvin, B.C., Kleinstein,
  R.N., Manny, R.E., Shin, J.A. and Mutti, D.O.,
  Myopia and ambient night-light lighting, Nature,
  vol. 404, pages 143-144, 2000.

11
CAUSALITY!

• How do you establish causality?
• what causes something else to occur?
• Lazarsfeld provided three criteria
• 1. Covariation two are related
• 2. Time order cause before outcome
• 3. Elimination of likely third variables

12
A classic experiment

• Independent and Dependent variables
• Cause and outcome
• Experimental and Control groups
• Exposed and non-exposed
• Pre-testing and Post-testing
• measure before and after
• "Blind" to condition
• placebo effects

13
Duct tape and warts

• Researchers at Madigan Army Medical Center put
  duct tape to the test against cryotherapy on 51
  patients, ages 3 to 22, who had come to clinics
  to be treated for warts.
• The patients were randomly assigned to receive
  either cryotherapy or duct tape. Patients in the
  cryotherapy group received a standard 10 second
  application of liquid nitrogen, which was
  repeated every two to three weeks. A small piece
  of duct tape was placed on the patients' warts in
  the other group. The patients were told to keep
  the tape on for six days. At the end of six days,
  they removed the tape, soaked the area in water,
  and then used an emery board or pumice stone to
  rub the dead skin off of the wart.
• Warts were successfully removed in 85 percent (22
  of 26) of the patients in the duct tape group,
  compared with 60 percent (15 of 25) of the
  patients in the cryotherapy group. The duct tape
  also caused fewer side effects than cryotherapy,
  resulting in little more than skin irritation. In
  comparison, the cryotherapy group reported pain
  and burning at the site.
• Overall, the researchers found duct tape to be
  more effective, less painful, and less expensive
  than freezing.

14
Test Units

• Subjects or entities whose response to the
  experimental treatment are measured or observed.

15
Establishing Control
16
Demand Characteristics

• Guinea pig effect
• Hawthorne effect

17
Experimental design

• Basic
• O- X - O experimental group
• O - . - O control group

18
Experimental design

• Solomon 4-group
• O- X - O experimental group (pretest-posttest)
• O - . - O control group (pretest-posttest)
• . X - O posttest-only experimental group
• . . - O posttest-only experimental group
• attempts to assess sensitization
• (from pretesting)

19
Experimental design

• Factorial designs
• Multiple variables are examined
• Number of factors
• One factor (Price)
• Two factors (Price and Packaging)
• Number of levels for each factor
• e.g., 2 by 2 design
• or 2 (Price) by 2 (Packaging) design

20
Field versus Laboratory Experiments
21
"Natural" experiments

• Naturalistic Quasi experiments
• measure effects of "naturally" occurring events
• advantages
• time ordering
• disadvantages
• cannot completely eliminate 3rd variables

22
Example Cause-Related Marketing
23
Validity

• External validity
• How generalizable are the findings?
• artificiality of the situation
• sampling issues

• Internal validity
• How confident are you in the relationships?
• this is the strength of experiments

24
Internal Validity

• History effect event related
• Maturation effect change in subject
• Testing effect pre-test education
• Instrumentation effect change in wording, etc
• Selection effect sample selection error
• Mortality effect subjects withdrawing

25
The End