Producing Data: Experiments

1
Chapter 9

• Producing Data Experiments

2
Experimentation

• Recall the distinction between experimental
  designs and observational designs
• In experimental studies, the investigator exposes
  individuals to a treatment to ascertain its
  effects

3
Vocabulary

• Subjects individuals participating in an
  experiment
• Factors specific experimental conditions or
  interventions applied to subjects
• Treatment a combination of a specific set of
  factors

4
Example Effects of Advertising

• Undergraduate students viewed a 40-minute video
  program that included ads for a digital camera
• Two explanatory variables (factors)
• Message length 30-second vs. 90-second
• Repetition commercial shown 1, 3, or 5 times
• Three response variables
• recall of the ads after viewing
• attitude toward the camera
• intention to purchase

5
Illustrative Example Treatments

• Factor A length of the commercial (2 levels)
• Factor B Number of repetitions (3 levels)
• Thus 2 3 6 treatments

6
Comparison
Comparison is first principle of experimentation
The effects of a treatment can be judged only in
relation to what would happen in its absence

• You cannot assess the effects of a treatment
  without a comparison group because
• Many factors contribute to a response
• Conditions change on their own over time
• The placebo effect and other passive intervention
  effects are operative

7
Randomization
Randomization is the second principle of
experimentation

• Randomization use of chance mechanisms to
  assign treatments
• Randomization balances lurking variables among
  treatments groups, mitigating confounding by
  lurking variables!

8
Blinding
Blinding is the third principle of experimentation

• Blinding assessment of the response in subjects
  is made without knowledge of which treatment they
  are receiving
• Single blinding subjects are unaware of
  treatment group
• Double blinding subjects and investigators are
  blinded

9
Illustrative Example Quitting Smoking with
Nicotine Patches

• Explanatory variable Nicotine patch / placebo
  patch
• 60 subjects, 30 assigned to each treatment group
• Response variable Cessation of smoking (yes/no)
• Design outline

Group 130 smokers
Treatment 1 Nicotine Patch
Random Assignment
CompareCessation rates
Treatment 2 Placebo Patch
Group 230 smokers
Source JAMA, Feb. 23, 1994, pp. 595-600
10
Randomizing Method

• Number subjects 01,,60
• Use table of random digits (TABLE B)
• Select a line arbitrarily (e.g.,
  line102)73676 47150 99400 01927
• First four subjects are 50, 40, 19, and 27
• Keep using table until you get 30 subjects in
  Group 1
• The remaining subjects are assigned to Group 2

11
Illustrative Example Mozart, Relaxation and
Performance on Spatial Tasks (Nature, 10/14/93,
p. 611)

• Subjects (30 undergraduate students) randomly
  assigned to one of three treatment groups
• Group 1 Listen to Mozart
• Group 2 Listen to relaxation tapes
• Group 3 Silence
• Response variable change in IQ score

Group 110 students
Treatment 1 Mozart
Random Assignment
CompareChange in IQ score
Group 210 students
Treatment 2Relaxation
Group 310 students
Treatment 3Silence
12
The Logic of Randomization

• Randomization encourages lurking variables to
  distribute evenly among treatment groups
• Difference in the response at end of treatment
  are then due to either
• Treatment or
• Chance assignment of treatments
• If the observed difference is larger than what
  would be expected just by chance, we say the
  results are statistically significant