Ch13 Flashcards
Quasi-Experiments
researchers can’t necessarily randomly assign participants to the level of the IV (which also may be a quasi-IV)
types of quasi-experiments
- Independent groups quasi-experiments
- Repeated measures quasi-experiments
Non-equivalent control group design
an ind. group quasi-experiment design- at least one treatment group and one comparison group (separate participants) but they haven’t been randomly assigned
types of Non-equivalent control group design (ind groups)
- Non-equivalent control group posttest-only design
- Non-equivalent control group pretest/posttest design
Non-equivalent control group posttest-only design
(ind groups) not randomly assigned to groups, and tested only after exposure to one level of the IV or the other
Non-equivalent control group pretest/posttest design
(ind groups) not randomly assigned to groups, tested both before and after intervention
Repeated-Measures Quasi-experiments (+ what does it rely on)
- analogous to repeated measures designs in true experiments- all participants experience all levels of IV, IV not manipulated
- Relies on something already occurring, like an already scheduled event, new policy, etc (ex: food break and parole decisions)
types of repeated measure quasi-experiments
- Interrupted time series design
- Non-equivalent control group interrupted time series design
Interrupted time series design
- repeated measures quasi-experiment study that measures participants repeatedly on a DV before, during, and after an “interruption” caused by an event (ex:food break and parole decisions)
- Also can be in real experiments, with a manipulated IV
- In is one- measured ordinally (1st, 2nd, 3rd)
Non-equivalent control group interrupted time series design
combines interrupted times series (repeated measures) and non-equivalent control group (independent groups) designs- with no experimental control over either IV
Internal validity threats in quasi-experiments
- selection effects
- design confounds
- maturation effects
- regression to the mean
- attrition threats
- testing and instrumentation threats
- observer bias, demand characteristics, placebo effects
selection effects (how to help)
Unaccounted for differences between groups- only relevant to ind groups
- Matched groups can help
- Waitlist design- have half the group get surgery right away, half wait (true experiment)
design confounds ( + how to help)
Extraneous variables (vary systematically)
- Look into and rule out
maturation effect (+ help)
- pretest/posttest- an observed change could’ve emerged over time
- Comparison group
history threat (+ help)
External historical event happens to everyone at the same time
- Comparison group
- Selection-history threat if only applies to one group
regression to the mean (+ help)
Occurs when extreme finding caused by random factors, unlikely to happen again, gets less extreme over time
- pretest/posttest only, only if the initial score is extreme
- In experiments, random assignment prevents it
attrition threats (+help)
When people drop out systematically- pretest/posttest
- Missing values analysis- shows if drop-outs were systematic
Testing and instrumentation threats (+ help)
Occurs when measured more than once- if not measured the exact same way
- Testing threats- participants’ answers change b/c they’ve been tested before
- Comparison group
observer bias (observer bias, demand characteristics and placebo effect) + help
Human subjectivity
- experimenter’s expectations influence interpretation of results- both construct and internal validity threatened
- masked or double blind
Demand Characteristics (Observer bias, Demand Characteristics, and Placebo Effects)
human subjectivity
- participants guess what the study is about change behavior
placebo effect (Observer bias, Demand Characteristics, and Placebo Effects)
human subjectivity
when participants taking a placebo improve because they think they’re receiving a treatment
- Comparison group
- Hiding what is being measured? Double blind?
External validity of Quasi-experiments
Real-world settings can enhance external validity due likelihood patterns observed will generalize
ethics in quasi- experiments
Many of the questions addressed in quasi-experiments would be unethical to study experimentally/ manipulate variables
construct validity in quasi-experiments
Usually show excellent construct validity for the IV - participants are actually doing the thing being measured
- Ask how successfully the DV was measured- reliable and valid?
Statistical validity in quasi-experiments
ask how large group differences were (effect size) + if results were statistically significant
Quasi-Experiments vs Correlational Studies
- Both can use independent-groups design
- Neither use random assignment
- Neither* use manipulated variables
Quasi-Experiments vs Correlational Studies- ind groups design
Both can use independent-groups design
- Comparing two or more levels of an IV (without random assignment)in order to predict one or more DVs,
Quasi-Experiments vs Correlational Studies- Neither use random assignment
Neither use random assignment
- IV not manipulated in quasi-experiments- but use a variety of techniques to enhance internal validity
- Can obtain people by targeting key locations, seeking out
comparison groups provided by nature or public policy
- In correlational studies, simply select some people, measure two variables, and test the relationship
Quasi-Experiments vs Correlational Studies- Neither* use manipulated variables
*Quasi-experiments use techniques like “active selection”
- Can obtain people by targeting key locations, seeking out comparison groups provided by nature or public policy
Small-N Designs
a study of only a few individuals
- Sometimes only 1 participant- single-N
for external validity, what is important?
How sample is selected is more important than size- for external validity
- If a large effect size is expected, a small sample can detect it
Large-N vs small-N (participants)
Participants:
- large-N: grouped- all participants in each group combined, studied together
- small-N: each participant treated as separate experiment
- Repeated-measures design- how they respond to
systematically designed conditions
Large N vs small N- data
Data:
large-N: data represented as group avgs
small-N: data for each individual
balancing priorities in case study research
Experimental control, manipulation, and replication
how are case studies like quasi-experiments?
Just like quasi-experiments take advantages of natural events, small-N studies often take advantage of special medical cases under controlled conditions
disadvantages of small-N studies
- Internal validity- difficult to isolate what’s being studied
- external validity
-Participants in small-N studies may not represent general
population- One option: triangulate
triangulate
- Compare case studies result to research using other methods
- Variety of evidence supports a parsimonious theory
triangulate- combining results of single-N studies with other groups
Behavior-change studies in clinical settings
- Can use Small-N designs in clinical settings to learn whether their interventions work
- Often used in behavior analysis- a technique in which practitioners use use reinforcement principles to improve a client’s behavior
- Studies designed to modify the behaviors of certain individuals to produce a desired result
behavior analysis
a technique in which practitioners use use reinforcement principles to improve a client’s behavior
Stable baseline designs
Stable baseline designs: a researcher observes a behavior for an extended baseline period before intervention or treatment is introduced
- If baseline is stable and the target behavior changes after the treatment, its effective
- Internal validity- stable baseline allows us to rule out maturation, regression to the mean, or a history effect
Then we can use replication to further enhance causal claims
Expanded rehearsal
Expanded rehearsal: a memory technique used on a real alzheimers patient
(standard baseline design)
Multiple-baseline designs + ex
researchers stagger interventions across situations, times, or contexts
ex: Contact desensitization plus reinforcement (autistic child gets treat when moving closer to therapy dog)
Reversal designs (ABA design):
Reversal designs (ABA design): researchers usually observes a baseline of behavior before treatment, next behavior during treatment, then they remove the treatment to see if the behavior reverts back (reversal period)
- Appropriate mainly in situations where a treatment may not cause lasting change
types of Behavior-change studies in clinical settings
- Stable baseline designs
- Multiple-baseline designs
- Reversal designs (ABA design)
internal validity in small-N designs
Can be very high if study is carefully designed- behaviors measured repeatedly
external validity in small-N designs
Can be problematic depending on goals of study- generalization not always relevant
1)triangulate- combining results of single-N studies with other groups
2) can specify population they want to generalize- might limit to a particular subset
3) might not care about generilizing in this study- might still be useful if only to one person
construct validity in small-N designs
Can be very high if definitions and observations are precise
- Sometimes need multiple observers and check for interrator reliability in case one observer is biased or behavior is difficult to identify
statistical validity in small N
Not always relevant to small-N studies- do not typically use traditional statistics
- Still draw conclusions from data and should treat it appropriately
- Often visual summary such a graphs are strong enough