QuasiExperimental Designs

1
Quasi-Experimental Designs
2
Quasi-Experimental Designs

• Intermediate between correlational study and true
  experiment.
• More than a relationship between variables.
• Low internal validity cannot determine
  causality.
• In true experiment, IV is manipulated and
  subjects are randomly assigned to conditions.
• In quasi-experiments, IV is manipulated, but
  subjects are already part of a group based on
  pre-existing characteristics.

3
Nonmanipulated IV

• IV occurs naturally
• Participants are not randomly assigned to
  conditions.
• Compares performance between 2 or more groups
  based on pre-existing characteristics.
• Ex gender religion age smokers vs.
  nonsmokers high, medium or low cholesterol
  levels.
• Groups are not equivalent before treatment.
• Low internal validity we cannot conclude
  causality
• Nonmanipulated independent variable and measure a
  particular dependent variable.

4
Control group Nonequivalent group

• True experimental designs have an experimental
  group (treatment) and a control group (no
  treatment).
• Participants are randomly assigned to either
  condition.
• Quasi-experimental designs do not have a control
  group because there is no random assignment of
  participants to the conditions.
• The nonequivalent group serves as the comparison
  to the treatment group

5
Typical quasi-experimental design

• Select 2 groups based on pre-existing
  characteristics.
• Divide each group in half half of the
  participants in each group get the treatment and
  half do not.
• Compare performance with and without IV within
  each group and across groups.
• Disadvantage
• Pre-existing differences can confound results.

6
Nonequivalent group design
Age Males Females
Caffeine Yes NO
DV of anagrams solved
7
Nonequivalent group design
Age Young Old
Memory Test Recall Recognition
DV of words remembered
8
Single-Case Experimental Designs
9
Single case experimental designs

• Involves the study of only 1 participant (single
  case designs) or 2 or 3 participants (small- n
  designs)
• Often used in clinical settings.
• Do not allow for generalization.
• Allow for replications with different IV on the
  same participant or small-n designs.
• Do not compare means nor run statistical
  analyses.
• Assess how performance changes from one condition
  to another by graphing it.

10
Baseline measurement

• A measurement of behavior made under normal
  conditions (e.g., no IV is present) a control
  condition.
• Serves to compare the behavior as affected by the
  IV.
• Collect enough measures to achieve a stable
  pattern.

11
Representative Single-Case Experimental Designs

• Reversal Designs
• IV is introduced and removed one or more times.
  
• 1) A-B design
• - simplest of all designs
• - measure baseline behavior, apply treatment and
  compare behavior after treatment to baseline.
• - does not allow to establish cause-effect

12
A-B design
treatment
Behavior during/ after treatment
Behavior at Baseline
13
A-B-A design

• Baseline measurement
• Apply treatment
• Measure change in behavior (posttest 1)
• Remove treatment
• Behavior should go back to baseline
• (final assessment)

14
A-B-A design
treatment
Behavior with treatment
Behavior at Baseline
Remove treatment
Behavior back to Baseline
15
A-B-A-B design

• Baseline measurement
• Apply treatment
• Measure change in behavior (posttest 1)
• Remove treatment
• Behavior should go back to baseline
  (assessment)
• Apply treatment again
• Measure change in behavior (posttest 2)
• More ethical to end with treatment.

16
A-B-A-B design
Remove treatment
treatment
Behavior at Baseline
Behavior with treatment
treatment
Behavior with treatment
Behavior back to Baseline
17
(No Transcript)
18
Multiple-Baseline Designs

• Effects of IV are assessed across several
  participants, behaviors and situations.
• Control for confounds by introducing treatment at
  different times for different participants,
  behaviors and situations.

19
Multiple-baseline designs

• Multiple-baseline across participants
• Determine who has most stable baseline and
  introduce treatment to that subject first.
• Multiple-baseline across behaviors
• Determine most stable behavior and start with
  treatment on that behavior and then start on 2nd
  behavior.
• Multiple-baseline across situations
• Determine when behavior is occurring and tackle
  one situation at a time.