QuasiExperimental Designs

1
Quasi-Experimental Designs

• Chapter 10

2
Comparison with True Experiments

• Goal of determining causation
• Where the research is being conducted
• laboratory versus natural settings
• Intervention or treatment ? IV
• More control in true experiments
• Randomly assignment to the levels of IV
• Appropriate comparison -- Simplest, two groups
  treated exactly alike except for the IV

3
External Validity

• Some research (e.g., cognitive, physiological,
  theory development, etc.) the setting doesnt
  matter -- basic research
• Quasi-experiments in natural settings allows for
  increased generalizability critical to
  intervention research (applied research) --
  practical goals

4
Obstacles to True Experiments

• Access and permission to study
• Permission to randomly assign ethical issues,
  especially when talking about a treatment
• Could offer the treatment to the control group if
  found to be of benefit

5
Threats to Internal Validity

• History
• Events other than the treatment confound the
  effect of IV
• Different experiences between groups
• Can control variables that are potential
  confounds, but that is more difficult with
  quasi-experiments
• Maturation
• Participants change over time
• Challenge to determine if change due to IV,
  especially in pre- post-test designs
• Need appropriate control group to be more
  confident due to IV

6
History

• Testing
• Experience and familiarity with testing
  procedures can effect DV (pre- post-test designs)
• Instrumentation
• Changes over time due to instruments (including
  the researcher) used to measure
• Can control with clear operational definitions
  and/or calibrating machines regularly

7

• Statistical regression
• Regression towards the mean when select based on
  extreme scores
• Low scores will be inflated because initially
  depressed due to error (chance or other factors)
• High scores will be depressed because initially
  inflated due to error
• Selection
• Random assignment minimizes differences between
  groups or averages those differences
• More challenging in natural settings

8

• Subject mortality
• Attrition changes the equivalence between
  groups
• Could be due to the IV
• Interactions with selection
• Any of the other threats interacts with selection
• Changes in freshmen may be greater than changes
  in sophomores maturation and selection
• history and selection different effect on one
  group compared to other (e.g., principle changes)
• Instrumentation and selection (ceiling and floor
  effects)

9
And even more problems

• Contamination
• Communication between groups about the experiment
  can result in -- resentment, demoralization,
  rivalry (especially with intact groups),
  diffusion of treatments
• External validity
• Representativeness of sample
• Generalizing across persons, settings, and times
  we rarely use random sampling
• Could even be difference between extra credit
  versus required participation
• Replication is key
• Experimental expectancy effects and the Hawthorne
  effect

10
Specific Quasi-Experimental Designs

• Lack of randomization is hallmark
• Need to be careful with causal interpretations
• One-group Pre-test and post-test design O1 X O2
  no internal validity
• Non-equivalent control group measure pre- and
  post-test in two groups one gets treatment, the
  other is the control
• O1 X O2
• O1 O2

11
example

• Type of communication given to elderly adults in
  nursing home

12
Interrupted Time Series

• O1 O2 O3 O4 X O5 O6 O7 O8
• Only assess abrupt changes because of small
  natural fluctuations
• Interrupted time series with nonequivalent
  control group
• Can address differences between groups on pretest
  by computing an ANCOVA analysis of covariance

13
(No Transcript)