Chapter 8 Flashcards
Characteristics of Experimental Research
The researcher decides on and carries out the specific interventions (or treatments) for one or more groups of subjects
The researcher controls extraneous variables so that they don’t pose alternative explanations for the results of the experiment
The researcher controls subject variables (characteristics) so that they are equivalent in different groups
a
1
b
2
C
3
D
4
a
1
b
2
C
3
D
4
Extraneous variables can be controlled
by eliminating them or by keep them constant
Internal Validity
The extent to which the outcomes of an experiment can be attributed to the experimental conditions, and not to other extraneous or confounding variables
When alternative explanations of the findings exist, we say that there are threats to _______________________.
the internal validity of the study
For threats to internal validity to be “plausible” two conditions must be met: __________________ and ____________________.
The factor must influence the dependent variable
The factor must be different in amount or intensity across levels of the independent variable
History
Threat from uncontrolled events that affect the dependent variable
Statistical Regression
Threat from change of extreme scores to those closer to the mean. Can be controlled through the use of an equally extreme control group
External History
Something that happened during the intervention outside the context of the study
Selection (Subject Characteristics)
Threat from the characteristics of participants differing as a function of condition. This threat is much less likely when subjects have been randomly assigned to conditions.
Maturation
Threat from changes in participants over time
Pretesting
Threat from the effect of taking a pretest
Instrumentation
Threat from the unreliability or changes in measurement
Treatment Replication
Threat from insufficient replications of the treatment
Subject Attrition (Mortality)
Threat from loss of subjects
Statistical Regression
Threat from change of extreme scores to those closer to the mean. Can be controlled through the use of an equally extreme control group
Diffusion of Treatment
Threat from treatment effect on one group affecting other groups
External validity
refers to the extent to which the results of a study can be generalized to other subjects, interventions, procedures, and settings
In the real world, there is generally a negative correlation between ___________________ and ___________; that is, studies strong in ___________ tend to be weak in _______________ and vice versa
internal validity and external validity
internal validity, external validity
External validity can be used to refer to generalization across:
Subjects Situations or Contexts Time Interventions or Treatments Measures
Single-Group Posttest-Only Design
A single group gets an intervention followed by a posttest
No comparison is possible in this design
Single-Group Pretest-Posttest Design
A single group gets a pretest, then an intervention, and finally a posttest
Posttest scores can be compared to pretest scores
Non-Equivalent-Groups Posttest Only Design
Two groups receive posttests, but only one receives the intervention
Several groups receive posttests, but each receives a different intervention
Non-Equivalent-Groups Pretest-Posttest Design
These designs mimic experimental designs, but lack random assignment of subject to groups, thereby being susceptible to selection or subject characteristic threats to internal validity
Randomized-to-Groups Posttest-Only Designs
The randomized to groups designs are true experimental designs, because they use random assignment of subjects to groups
A control group receives no intervention; a comparison group receives the “standard” intervention
Random assignment produces groups that should be equivalent, as long as the N is large enough (generally at least 15 per group)
Sometimes called a quasi-experimental design due to the lack of random assignment
Non-Equivalent-Groups Pretest-Posttest Design
Subject Effects (Reactivity)
Threat from the effects of awareness of being a subject in a study. Examples are the Hawthorne effect, compensatory rivalry, resentful demoralization, and the novelty effect.
Generalization over the type of and nature of measures used to represent the dependent variable require ________________.
studies to support such generalization.
Interventions or Treatments
Generalization over different way of delivering a conceptually identical intervention requires _____________.
new studies.
Generalization over time periods generally requires _______________.
additional studies at the new time periods.
Generalization to new situations or contexts can be assessed through _________________________-.
sampling and statistics, but usually isn’t. Usually a logical argument is made by the experimenter.
Ecological Validity
is the term used for generalizing from the laboratory to the real world.
Generalization from the sample to sub-samples (e.g., different genders or ethnic groups) of the total population requires _____________.
additional statistical analyses
Experimenter Effects
Threat from characteristics or expectations of the experimenter. Characteristics can include age, sex, race, status, hostility, authoritarianism, and physical appearance. Experimenter expectancy refers to deliberate or unintentional effects of bias on the part of the experimenter, which is reflected in differential treat of participants. These can be eliminated by running the experiment double-blind.
Subject Effects (Reactivity)
Threat from the effects of awareness of being a subject in a study. Examples are the Hawthorne effect, compensatory rivalry, resentful demoralization, and the novelty effect.
Factorial Designs
Studies that employ two of more IVs, or interventions, each of which may have multiple levels
Field studies should check ________________.
intervention fidelity.
Criteria for Evaluating Experimental Research
The primary purpose is to test causal hypotheses
There should be direct control of the intervention
The experimental design should be clearly identified
The design should provide maximum control of extraneous/confounded variables
The intervention should be described and implemented as planned
The determination of n should be the same as independent replications
The measure of the dependent variable must be sufficiently sensitive to capture the change caused by the intervention
Single-Subject (Single Case) Designs
One or a few subjects are studied, rather than a large group
The subject is measured multiple times prior to and after an intervention
Behavior prior to the intervention is compared to behavior after the intervention to determine the effectiveness of the intervention
Baseline
The period of time in which the target behavior is observed without the intervention
Intervention
The period of time in which the target behavior is observed with the intervention
A-B-A Design
Sometimes called a withdrawal design
Characteristics of Single-Subject Research
Reliable Measurement Repeated Measurement Description of Conditions Baseline and Intervention Conditions Single-Variable Rule
Multiple-Baseline Designs
Observations are made on one of the following: Several Subjects, Several Target Behaviors in a Single Subject, Different Situations (or Contexts)
The onset of the intervention is offset in time for each of the: Subjects, Target Behaviors, Situations or Contexts
Criteria for Evaluating Single-Subject Research
There should be reliable measurement of the target behavior
The target behavior should be clearly defined
Sufficient measures are needed to establish stability in behavior
Procedures, subjects, and settings should be fully described
A single, standardized intervention should be used
Experimenter or observer effects should be controlled
Results should be practically significant
