How can we help?

You can also find more resources in our Help Center.

33 terms

Quasi-Experimental & Single Subject Designs

STUDY
PLAY
quasi-experiment
when the researcher can't assign participants randomly to conditions OR they are unwilling or unable to manipulate the causal variable of interest
lack of control
why do quasi-experiments have less internal validity than normal experiments?
internal validity
the degree to which a researcher draws accurate conclusions about the effects of an independent variable on participants' responses
low validity
studies lack necessary controls to draw any meaningful conclusions
high validity
studies in which the experimental design and tight control allow us to rule out every alternative explanation for the findings
external validity
to what other settings, groups, treatment variables, and measurement variables can an experimental outcome be generalized?
how to increase external validity
- move from the lab to the field
- this makes it less easy to rule out alternative explanations, though
single-group design
-BAD QUASI EXPERIMENTAL DESIGN

give a treatment and measure some behavior.

- there's NO RANDOM ASSIGNMENT!
- also, REACTIVITY
time-series design
within-subject designs -- performance of a single group is measured several times before and after introducing an experimental treatment
simple interrupted time-series design
a single group of participants in tested repeatedly before and after a manipulation or a natural event

- the multiple measures allow us to detect many confounding variables
program evaluation
research designed to assess the effects of interventions that supposed to influence behavior
practical problems with program evaluations
- the 'participants' are actually part of the program
- a no-treatment control group may raise ethical and political questions
- treatment staff may be threatened by the evaluation of irritated by the extra work load
control in program evaluation
- selecting appropriate dependent measures
- avoiding bias in the evaluation
- using strongest research design possible
single-subject design
- the unit of analysis is the individual participant
- each participant's responses are analyzed separately and the data are rarely averaged across participants
- YOU CAN'T DO INFERENTIAL STATISTICS

- the process under study is found within a single individual and can be controlled appropriately; thus a sample isn't necessary
error variance
first advantage of single-case over group designs

- much variance in group data is due to individual differences (bad variance). single-case studies can help to understand good variance
generalizability
second advantage of single-case over group designs


single-case researchers argue that group averages may not accurately portray the response of any particular individual. If so, (blank) from overall group may be mosleading
reliability
third advantage of single-case over group designs

single participant experiments replicate the effects of the independent variable
ABA (reversal) design
- measure baseline performance over one or more sessions
- give treatment over one or more sessions
- remove treatment over one or more sessions
- give treatment a second time over one or more sessions
multiple treatment (ABACA) design
- measure baseline performance (A) over one or more sessions
- give treatment B over one or more sessions
- remove treatment B over one or more sessions (go back to baseline A)
- give treatment C over one or more sessions
- - remove treatment C over one or more sessions (go back to baseline A)
will only work when you are studying the effect of treatment conditions on behaviors that return relatively quickly to baseline levels
what is the limitation of reversal designs
multiple baseline designs
- two or more behaviors are studied simultaneously
- after you get baselines on all the behaviors, an IV is introduced that is hypothesized to affect only one of the behaviors
- the selective effects of a variable on a specific behavior are then documented
- try to show that the IV caused the target behavior to change but did not affect the other behaviors (correlation would be bad)
- then repeat with the next behavior
graphs that show the results individually for each participant
what is the preferred method of presenting data from single-participant designs?
comparison group
a group that is expected to be similar but not equivalent to the experimental group (random assignment has not been used)
comparison-group design
research that uses more than one group of individuals that differ in terms of whether they have had or have not had the experience of interest
selection threats
threats to internal validity that occur because individuals select themselves into groups rather than being randomly assigned to groups
single-group before-after design
-BAD QUASI EXPERIMENTAL DESIGN

research that uses a single group of participants who are measured before and after they have had the experience of interest

- there's NO RANDOM ASSIGNMENT!
- also, REACTIVITY
retesting threats
potential threat to internal validity. whenever DV's are measured more than once, participants may be able to guess the research hypothesis.
attrition
potential threat to internal validity in longitudinal studies. students who stay in the program may be different than those who drop out
maturation threats
potential threat to internal validity. changes in participants over time but not due to the IV.
history threats
potential threat to internal validity. occur due to the influence of changes in the social climate during the course of the study.
comparison-group before-after design
research in which more than one group of individuals is studied and the dependent measure is assessed for all groups before and after the intervening event

- allows the scientist to control for some of the threats to validity that occur in before-after studies that last over a period of time
regression to the mean
a statistical artifact such that whenever the same variable is measure more than once, if the correlation between the two measures is less than r=1.00 or greater than r= -1.00, then the individuals will tend to score more toward the average score of the group on the second measure than they did on the first measure, even if nothing has changed between the two measures
participant-variable design
a research design in which one of the variables represents measured differences among the research participants, such as demographic characteristics or personality