Wed Oct 29, 2003

1
Wed Oct 29, 2003

• Exams after class
• Mean(SD) 71.6(11.6)
• Range 44-95
• Appointment
• Research day
• 1 week

2
Distribution
3
Quasi-experimental

• Have one or more
• manipulated variables (X)
• Outcome variables (O)
• No random assignment
• Pre-assigned groups
• Ethical or practical reasons
• CHD vs. No-CHD
• Smoker vs. non-smoker
• Better example CR vs no-CR

4
Quasi-experimental
O X O --------- O O

• Non-equivalent control group design
• Better than previous designs
• Control group, pre post-tests
• More confidence in causal inferences

5
Problems with this design

• Selection bias
• Most participants self-selected
• Pre-existing groups
• Can affect pre-test scores
• Maturation
• History
• regression

6
Interpreting outcomes

• Outcome 1
• Selection threat
• Maturation?
• History?
• Treatment group
• Not control

7
Interpreting outcomes

• Outcome 2
• Selection
• History
• Maturation

8
Interpreting outcomes

• Outcome 3
• Selection
• Large decline?
• Regression
• Example?

9
Interpreting outcomes

• Outcome 4
• Selection
• Large increase?
• Regression
• Example?

10
Interpreting outcomes

• Outcome 5
• Clearest evidence of successful intervention
• Unlikely to change faster with lower baseline

11
Quasi-experimental
O1 O2 O3 X O4 O5 O6

• Interrupted time-series
• Similar to pre post test design
• Multiple pre post-test measures
• Increased confidence in baseline and outcome data
• Selection, maturation, history

12
Outcomes of interrupted
Treatment
13
Quasi-experimental
O1 O2 O3 X O4 O5 O6 X O7 O8 O9

• Interrupted time-series with removal
• To show change with X and change with removal of
  X
• e.g., drug regimen, knee brace, etc.
• Question of ethics (removal of X)
• Should be a quick treatment response

14
Quasi-experimental
O1 O2 O3 X O4 O5 O6 -------------------------- O1
O2 O3 O4 O5 O6

• Interrupted time-series with non-equivalent
  control
• Stronger that non-equivalent control
• More confidence in pre post-test measures
• Strongest Quasi design?