Stat 100

1
Stat 100

• Jan. 25

2
To Do

• Read Chapter 5

3
Key Terms

• Observational Study essentially a survey,
  investigator does not assign any tasks to
  participants
• Randomized Experiment researcher assigns
  different treatments, tasks or conditions to
  participants

4
Experiment or observational study?

• 100 smokers who want to quit are randomly divided
  into two groups.
• Researchers give nicotine patches to one group
  and placebo patches to the other.
• Percents not smoking after two months are
  compared.
• Its an experiment treatments are assigned
  randomly to the participants

5
Randomization

• Randomly dividing participants into treatment
  groups
• Why randomize?
• Reduces risk of biased comparison. Random
  assignment to groups should provide comparable
  groups.
• Groups should have about same characteristics in
  the beginning.

6
Nicotine patch example

• Randomly assigned treatment groups should have
  about the same motivation, same degree of
  addiction, etc.

7
Placebo

• Fake treatment
• Why use a placebo?
• In nicotine patch example, some will quit even
  without a treatment.
• Use of placebo allows researchers to better
  assess how much better nicotine patch is than
  doing nothing

8
Placebo Effect

• The observed effect that some people get better
  even if they use a placebo.
• In past, attributed to psychological factors.
• Another explanation some get better even
  without treatment
• Existence of psychological placebo effect under
  debate.

9
Ethics of using a placebo

• Not always ethical, so not always used
• Comparison is always desirable, so a control
  group often given the existing standard treatment
  when a new treatment is evaluated

10
Blind, double blind

• Blind either participant or evaluator doesnt
  know what treatment is used
• Double blind neither participant nor evaluator
  knows the treatment assignment
• Nicotine patch surely was blind. Participant
  wouldnt know if placebo or nicotine patch is
  used.

11
Hawthorne effect

• Often, results in an experiment are better than
  what will occur in real world use of treatments
• For instance, more might quit smoking in
  experiment than in normal real world setting
• Why? participation in experiment might provide
  additional motivation

12
Example

• Experiment is done to assess effect of alcohol
  consumption on driving ability
• Different groups consume different amounts of
  alcohol
• For safety reasons, driving test is done using a
  driving simulator
• What difficulties does this present for
  interpreting results?

13
Ecological Validity

• Conditions of experiment should resemble real
  world conditions
• Alcohol and driving ability experiment might lack
  ecological validity
• Because simulated, not actual driving is done.

14
Alcohol and driving, continued

• Suppose there are 3 alcohol conditions none,
  moderate amount, high amount
• How should participants be assigned to conditions?

15
Assignment methods

• Only randomized randomly divide participants
  into three groups, each assigned a different
  alcohol condition
• Blocked find three similar individuals,
  randomize one to each condition, repeat this idea
  over all participants

16
Criterion for Similar

• In driving and alcohol experiment, what variables
  would be used to find similar people?
• What factors affect driving and response to
  alcohol?
• Weight, age, gender, drinking habits

17
Blocks

• Find three people who are roughly same age,
  weight, and gender(?).
• Assign one to each condition
• Repeat this idea over all participants

18
Why block?

• Provides more direct control over the
  comparability of treatment groups.

19
Observational studies

• Covered on Jan. 9, See those notes
• Note PowerPoint slides are at course web site
• Also, see Ch. 5