Chapter 12: Quasi-Experimental Designs

1
Chapter 12 Quasi-Experimental Designs

• When researchers can not manipulate the
  independent variable, rather it is a grouping
  variable (gender, age, disability) and
  equivalence between the groups can not ensured
• Researchers can not randomly assign participants
  to groups thus lack control over extraneous
  variables
• Quasi-independent variable is not a true
  independent variable but usually occurs naturally
  or can not be manipulated.
• Researchers still look for effect of the
  quasi-independent variable.

2

• Quasi-experimental designs usually have lower
  internal validity than true experiments.
• Types of Quasi-Experimental Designs
• Pretest-Postest designs
• Test participants before an after the
  quasi-independent variable
• One group measure participants before and after
  the quasi-independent variable. Only have one
  group of participants (those that experienced the
  quasi-independent variable)
• Test reading before children at school X start
  reading program and then test their reading after
  they finish the reading program.
• O1 X O2

3

• Threats to internal validity
• Maturation students may have matured over the
  reading program. They may be better at reading
  just because of time and not due to the program.
• History Effects something other than the
  independent variable may have occurred between
  the pretest and posttest.
• Pretest sensitization taking the pretest may
  change the participants reaction to the posttest.
• Regression to the mean Tendency for extreme
  scores on pretest to regress (move towards) the
  mean on a subsequent test (posttest).

4

• If participants are selected because they have
  extreme scores on the pretest (e.g. select a
  school with very poor reading ability) there may
  be other factors due to measurement error that
  resulted in such low scores at the pretest
  (tired, bad day etc.) that may have slightly
  deflated their scores.
• Measurement error causes extreme scores to be
  biased in the extreme direction (away form the
  mean).
• So when you test them a second time it is
  unlikely that you will have those same factors
  that may have deflated their scores and their
  scores will increase a bit and make it look like
  the program is have an effect.

5
(No Transcript)
6

• Nonequivalent Control Group Design
• We cannot randomly assign participants to control
  and study group, so we select a control group
  that is similar to the group that gets the
  quasi-independent variable.
• Posttest-only measure both groups after one
  group has received the treatment.
• Measure reading in School A and School B after
  School A has participated in the reading program.
• Quasi-experimental group X O
• Nonequivalent control group -- O
• Selection bias we do not know whether the two
  groups were similar before the intervention

7

• Pretest-Posttest design Test both groups before
  one group gets the intervention, then test both
  groups again after one group gets the
  intervention (quasi-independent variable)
• Quasi-experimental group O1 X O2
• Nonequivalent control group O1 -- O2
• Allows researchers to see if the two groups
  scored similarly on the dependent variable before
  the introduction of the treatment.
• To determine if the quasi-independent variable
  had an effect you want scores to change between
  pretest and posttest ONLY for the
  Quasi-experimental group and NOT for the
  Nonequivalent control group.

8

• Time Series Designs
• Measure the dependent variable many times before
  and after the quasi-independent variable is
  introduced.
• Simple interrupted time series design
• Researchers make a series of observations of the
  dependent variable before and after the treatment
  is introduced.
• O1 O2 O3 O4 X O5 O6 O7 O8
• Evidence for a treatment effect occurs when there
  are abrupt changes in the time-series data at the
  time the treatment was implemented.

9
Percentage of cavities after the introduction of
fluoride into toothpaste in 1970
10
Percentage of cavities after the introduction of
fluoride into toothpaste in 1970
11
Percentage of cavities after the introduction of
fluoride into toothpaste in 1970
12

• This design helps to distinguish changes due to
  maturation from the quasi-independent variable
• Contemporary History Observed effect could still
  be due to another event that occurred at the same
  time as the quasi-independent variable
• Perhaps the electric toothbrush was introduced in
  1970, or there was a major TV add campaign that
  promoted brushing teeth.

13

• Interrupted time series with a reversal
• Researchers measures the dependent variable
  before and after the treatment is introduced and
  then again after the treatment is removed
• O1 O2 O3 O4 X O5 O6 O7 O8 -X O9 O10
  O11
• We can see what happens to the dependent variable
  after the quasi independent variable is
  introduced and then again after it is removed.
• If the quasi-independent variable was really
  having an effect we would expect performance to
  change back to normal after it is removed

14
Percentage of cavities after the introduction of
fluoride into toothpaste in 1970
15

• Limitations
• Researchers may not have the ability to remove
  the quasi-independent variable
• remove fluoride from toothpaste, remove a
  seatbelt law
• Some effects of the quasi-independent variable
  may remain even after it is removed
• If you did a time series study before and after
  introduction of reading program, and then removed
  program, reading may not decrease, the children
  may not regress because they did learn to read.
• Removal of the quasi-independent variable may
  produce effects that are not due to the
  quasi-independent variable

16

• Control Group Interrupted time series
• Measure more than one group on several occasions,
  but only one group receives the quasi-independent
  variable.
• O1 O2 O3 O4 X O5 O6 O7 O8
• O1 O2 O3 O4 -- O5 O6 O7 O8
• Helps to rule out history effects, and we can be
  more certain the a change was due to X rather
  than an outside influence.
• Could still have a local history effect.

17

• Longitudinal Designs
• Time serves as the quasi-independent variable
• Commonly used in developmental research
• Allows researchers to eliminate generational
  effects (when effects differ depending on the era
  in which people grew up). In longitudinal
  research you are studying people of the same
  generation over time.
• O1 O2 O3 O4 O5 O6 O7 O8
• Allows researchers to examine how individuals
  change with age (not just group differences).

18

• Limitations
• Difficult to obtain samples who are willing to be
  repeatedly tested over time.
• Difficult to keep track of participants over time
  (attrition may occur).
• Takes a lot of time and money.

19

• Program Evaluation
• Used to assess effectiveness of interventions (or
  programs) and provide feedback to the
  administrators
• Assess needs, process, outcome, and efficiency of
  social services.
• Considered applied research.

20

• Evaluating Quasi-Experimental Designs
• The presumed cause usually precedes the presumed
  effect.
• These designs do allow researchers to determine
  if the two variables covary together.
• BUT they can not eliminate effects of extraneous
  variables and ensure randomization.

21
Chapter 13 Single-Case Research

• Examine individual participants rather than a
  group of participants.
• Idiographic approach describe, analyze, and
  compare individual behavior
• Nomothetic approach Describe, analyze, and
  compare behavior across individuals and make
  broad generalizations to a group.
• Many research areas started in single
  case-research (Ebbinhgauss memory research on
  himself, Skinner and Pavlov research)

22

• Three main criticisms of group designs
• Error variance error variance in group data does
  not always reflect variability in behavior
  (rather is due to the design).
• Group designs examine inter-participant variance
  which is across or between participants
  (individual differences)
• Rather single-case researchers emphasize
  intra-participant variance which is variability
  in an individuals behavior.
• Generalizability single case researchers suggest
  that group designs have limited generalizability.

23

• Group designs usually reflect an average of the
  participants behavior which may not represent the
  response of any particular participant.
• Average number of children adults have 2.1
• Average anxiety score is 10 (but most people may
  be at either high or low end 3-4 and 17-18).
• Reliability group designs may test an effect
  once, but do not always replicate it to see if
  the effect holds up and is reliable.
• Single-case researchers often test an effect in
  the same participant a few times
  (intraparticipant replication) or determine
  whether the same effect is found in a few other
  participants (interparticipant replication)

24

• Single-Case Experimental Designs
• ABA design observe participants in absence of
  independent variable, baseline (A), then
  introduce independent variable, experimental
  period (B), then remove independent variable and
  observe behavior (A).
• Sometimes called a reversal design
• Difficult to determine if some other event
  occurring at the same time as the independent
  variable resulted in the effect
• The independent variable may produce permanent
  changes in a participants behavior, so it may not
  go back to baseline.

25

• Multiple-I Designs
• Present varying nonzero levels of the independent
  variable
• ABC design baseline (A), introduce IV (B), then
  remove this IV and introduce another level of the
  IV (C).
• ABACA have baseline condition between each level
  of the independent variable
• Multiple Baseline Designs two or more behaviors
  are examined simultaneously.
• After the baseline data, the researcher examines
  the effect of the independent variable on both
  behaviors. Usually test to see if independent
  variable affects the hypothesized behavior.

26

• Data Analysis
• Graphic analysis display all of a participants
  data points (before and after independent
  variable) on a graph. Visually examine the graph
  to determine whether it looks like the
  independent variable produces an effect.
• Uses of single-case designs
• Conditioning research (reinforcement and
  punishment effect)
• Behavior modification techniques (phobias)
• Demonstrational purposes.

27

• Critiques of single-case research
• Not necessarily generalizable because the
  participant is not usually chosen at random
• Difficult to study interactions among variables
• Ethical issues in ABA designs when the researcher
  may remove a very helpful treatment

28

• Case Study Research
• An intensive description and analysis of a single
  individual, or sometimes a group
• Usually gather a narrative description of
  information about the individual(s).
• Common when describing rare phenomena (rare brain
  injuries or disorders, prodigies, assassins)
• Psychobiography use psychological theories to
  understand lived of famous people (study Nixon)
• Used to make anecdotes and to illustrate general
  principles.

29

• Limitations of Case Studies
• Very difficult to control extraneous variables.
  Usually unable to asses and rule out alternative
  explanations.
• Observer Biases