QuasiExperimental Designs

1
Quasi-Experimental Designs
2
Quasi-Experimental Design

• You cannot randomly assign experimental
  participants to groups
• You DO manipulate an IV
• You DO measure a DV

3
When to use quasi-experimental design?

• Study participants in certain groups
• Evaluate an ongoing or completed
  program/intervention
• Study social conditions (examples poverty, race,
  unemployment)
• Expense, time, or monitoring difficulties
• Ethical considerations

4
Nonequivalent Groups Design

• Structured like a pretest-postest randomized
  experiment
• Key is to create as equal a comparison group as
  possible through our selection criteria

N O X O (treatment) N O O (comparison)
5
Nonequivalent Groups Design

• Example Geronimous (1991)
• Typical outcomes for teen mothers
• Poverty, high-school dropout rates increase,
  higher infant mortality
• Geronimous believed that family factors, like
  SES, were better predictors of outcomes than teen
  pregnancy
• How to find a comparison group as similar as
  possible?

6
Nonequivalent Groups Design Analysis

• Your groups began the experiment as not
  equivalent
• So, the important question is not whether there
  was a difference
• Is the difference the same as before the
  experiment?

7
(No Transcript)
8
(No Transcript)
9
(No Transcript)
10
(No Transcript)
11
(No Transcript)
12
(No Transcript)
13
Nonequivalent Groups Design

• Threats to Internal Validity
• Maturation
• Instrumentation
• Statistical regression
• Interaction between selection and history
• Interpretation of findings must be more cautious,
  but a strong research design

14
Interrupted Time-Series Design

• Measure a group of participants repeatedly over
  time
• Interrupt with a treatment
• Measure participants repeatedly again

O1 O2 O3 O4 X O5 O6 O7 O8
15
(No Transcript)
16
(No Transcript)
17
Interrupted Time-Series Design

• Threats to Internal Validity
• History
• Maturation
• Instrumentation

18
Interrupted Time-Series Design

• How to control for history threats?
• Frequent measurement intervals
• Comparison group

19
Interrupted Time-Series Design

• How to control for history threats?
• Frequent measurement intervals
• Comparison group
• Measure, treatment, measure, then undo the
  treatment, measure again
• Not always feasible