5- fundamentals of single case experimental design l Flashcards
- Small number of subjects
- multiple, repeated observations
- When independent variable is introduced, changes in the independent variable are made once the dependent variable has reached a steady state
Single case designs
The variable in an experiment measured to determine if it changes as a result of manipulations of the independent variable. In ABA, it represents some measure of socially significant behavior.
The variable that is systematically manipulated by the researcher in an experiment to see whether changes in the independent variable produce reliable changes in the dependent variable. In ABA, it is usually an environmental event or condition antecedent or consequent to the dependent variable.
Dependent variable
IV introduction:
Changes in the IV assigned according to randomize and match designs.
Method of data analysis: Usually inferential statistics
Exposure to the treatment:
Each individual often exposed to only one level of the IV….That is, either baseline OR treatment
Comparison is made between groups of individuals
Control group versus experimental group.
Large numbers of subjects.
Few, often single observations
Group design:
- Permit investigation of behavior change as a dynamic process
- Repeated measurements and stability criteria mean that you need to keep observing
- Permits you to see patterns of behavior change
- Taking single measurements obscures that process
Single case design advantages: dynamic change. …Relative to group designs
Allows examination of interest subject variability
Group Design may not reveal anything about the performance of any given individual
eg, and increased group score might reflect great improvement for a small number of subjects, while the majority did not change or performance decreased
Allows the examination of intra- subject variability
Difficult to detect sources of within subject variability with group data
Better for exploration of idiosyncratic affects and serendipitous findings because the data path is not constrained by hypothesis testing
Single group design Advantages : Variability
The goal is to demonstrate a functional relation between the independent and dependent variables
- Functional relation: when A, and only A, causes B
- Determined/demonstrated through systematic manipulation
The repeated and systematic presentation and removal of an independent variable E.g.,The intervention while measuring the dependent variable(s) (Target bx) , and holding other factors constant
There is no distinction in systematic manipulation used for treatment of evaluation for experimental design
Experimental design in ABA
provides information about existing extent of the problem.
Serves as an indication of whether the intervention is necessary.
Baseline function
Movement from one kind of independent variable to another kind of independent variable describes what?
gather baseline data for three sessions
Need at least three to start a trend.
Regardless of length, consider the last three points as a small trend.
A qualitative research for determining if the trend is sufficiently stable
Example: no more than 25% deviation in last three points. No trend in any direction in last three points.
Particularly important in research
When in doubt, Run it out
Phase change
You are making tremendous progress with a research project when suddenly the behavior dramatically changes. After interviews you were informed that the subjects parents divorced. What is the probable threat to internal validity?
History
I am going to do a withdrawal design with a four-year-old who receives frequent praise for appropriate replacement behaviors. During the with drawl face his mother continues to give him praise occasionally.
Diffusion of treatment
I am running a long-term study on the effects of a language acquisition program. I work with children from the time they are one and a half until they are four years old. All of my participants increase their language.
Maturation
I am doing a study with group home staff on increasing performance. One week into my study the group home manager quits.
History
I am examining the effects of a new teaching program on receptive identification. Aside from my intervention, one of my participants is also receiving outside therapy on identified picture cards.
History
I am doing a study on reducing tantrum behavior, and I still have interobserver agreement of 50%.
Instrumentation
Based on the data above what is your conclusion about the response class program?
It has a reductive effect on tantrums
Assessment of the dependent variable prior to the introduction or change of the independent variable
Does not necessarily imply the absence of treatment
State of events before you made any changes
Baseline logic BL Definition
Baseline ohase A: independent variable is absent
Intervention phase B: independent variable is introduced and remains present
Baseline data are first collected on two separate behaviors, versus withdrawal, and what you think of behavior is tracked.
Next, treatment one is applied to behavior one; at the same time, treatment to his apply to behavior to.
Treatment 2 is usually the reversal I have treatment one, but can be a no treatment a.k.a. BL condition
Return to baseline A: independent variable is withdrawn
If behavior changes systematically as a function of the introduction and withdrawl of the independent variable,
The likelihood is small that some extraneous variable produced the behavior change.
This likelihood decreases with each subsequent withdraw an introduction of independent variable Prediction, verification, and replication
instead of simply withdrawing treatment one from behavior 1 In the next phase, the treatments on the two behaviors are reversed. Behavior one receives treatment to, and behavior to now receives treatment one.
In the final phase, the interventions are once again reversed: behavior number one again receives treatment number one, and behavior two goes back to receiving treatment two
Essentially this is two concurrent but out of phase with drawl designs, imposed on two behaviors, which are possibly functionally related.
Reversal designs: ABA
Reversibility – some behavior changes are not reversible:
Intervention produces effects that are impossible to withdraw e.g., skills acquisition
Behavior after initial change, makes contact with other variables that make reversal unlikely even intervention is withdrawn
E.g., contrived reinforcement to support Social initiations – initiations get support by the response of others
The ethics of intervention reversal
In cases, it may be an ethical to reverse treatment
You’re going to do what now with my child?
Balancing reversal ethics with the right to effective treatment
It may be unethical not to show a functional relation
Which phase first- does the analysis always start with the absence of the independent variable?
If intervention is immediately critical, Analysis may begin with an intervention phase
Beginning with intervention does not alter the logic of the design
Most straightforward single case arrangement
Most powerful demonstration of functional relationships
Reverse limitations : Irreversibility
Ethics
This distinction was originally going to be made by the fourth edition task list, but since it is no longer commonly used either in practice of literature, the distinction is no longer required by the board
Withdrawal versus Reversal
Baseline is followed by a treatment condition. The intervention used in the treatment condition is then withdrawn and returned to baseline.
Baseline, BL, and a treatment I, Tx, or alternated,
A-B-A design
Often ends in treatment: A B A B
Withdrawal
Three standard variations of multiple baseline data
Multiple baseline across
subjects
behaviors
settings
When two or more base lines are established, followed by the staggered introduction of the independent variable to each baseline, this procedure describes which experimental designs.?
Inappropriate when the behaviors are interdependent
Changing behaviors and one baseline is likely to change behaviors and the other baselines even before the independent variable is presented
Intermittent measures, or probes are taken rather than continue with measurement on each baseline
Procedurally
Two or more independent baselines Are established.
The independent variable is then separately introduced in a staggered fashion to each
When baseline is stable for the first baseline, the independent variable is introduced on the second baseline and so on (design logic)
Experimental control is demonstrated by showing that behavior changes when, and only when, the independent variable is introduced to each baseline
The possibility of extraneous variables causing the change is highly unlikely under the circumstances
Usually dictated by the stability, level, and trend of the first baseline
Same rules apply, as long as necessary, as short as possible
Often first baseline is continuous, but subsequent baseline data collection is conducted on an intermittent basis relative to the first base
How many baselines? The larger the number, the more convincing. Lanes both internal and extra no validity
Using only two base lines in a multiple baseline design can be a risk.
If one does not change, the conclusions are questionable, essentially an AB design
If 3+ our use and one does not change, still a reasonable demonstration of experimental control, the failure is the likely outlier
Multiple baseline design
One of the disadvantages of the multiple baseline design is that the experimenter cannot test Methods and interventions before applying them on a larger scale. True or false
False
A.k.a. delayed multiple baseline
Separate baselines are taken
The baselines are staggered in terms of the number of data points in the same way
But the baselines are not conducted at the same time
Advantage
Permits greater flexibility in the analysis – not constrained by having to have all subjects concurrently present
Limitations
Analysis need not be completed concurrently because the chances that the same extraneous variable, rather than that IV reduced the same result is even more remote if the baselines are separated by time
Non-concurrent multiple baseline. Design
Which is not an advantage of the multiple prop technique.
A. Less time effort to implemen
B. Avoid threats stemming from repeated exposure such as extended practice
C reduces likelihood of variability
C
- Involves:
Repeated, systematic presentation and removal of an independent variable
Changes in environmental events, interventions
While measuring changes in the dependent variable:
Some aspect of the behavior of interest
And holding other factors constant
- Goals:
to demonstrate functional relation between the independent variable and the dependent variableEvaluate the interventions once they are decided upon
- Experimental design.
2. Experimental design: goals
In ABA there is no distinction between:
Systematic manipulation used to evaluate The effects of a treatment
AND
Systematic manipulation used to answer a Research question
Experimental design: research and practice
When changes in an antecedent or consequent stimulus class consistently alter a dimension of a response class
Hence, A functional analysis -
Quantitative direct observation of behavior under systematically manipulated and controlled conditions;
-NOT Just a method of assessing problem behavior
Functional relation
Quantitative direct observation of behavior under systematically manipulated and controlled conditions; not just a method of assessing problem behavior
Functional analysis
The extent to which an analysis assures us that measured changes in behavior Are due to the manipulation and not due to uncontrolled extraneous variables
Primacy of Internal Validity
Regarded as a PRIORITY over external validity
It makes little sense to consider the generality of the effect If EXTRANEOUS VARIABLES cannot be rolled out for the effect
Internal validity
The extent to which a study’s results are generalizable to other subjects, settings, or behaviors
External validity
History
introduction of the independent variable may coincide with other events in the persons life.
Those other events could have produced those effects
Maturation Testing Instrumentation Diffusion Threat regression Selection bias Attrition
Internal validity threats
Inadvertent, uncontrolled seepage of the treatment to control conditions or control subjects
E.g., parent gets child to practice new skills before the intervention is formally introduced
Diffusion threat
Changes occurred before baseline measurements were not representative of the natural state of events. E.g., unusual events took place on the initial day of testing which were not in place after intervention, so it looks like the intervention was effective.
Perhaps less relevant to single case designs because of repeated measurement
Regression: internal validity threat
Changes in the dependent variable may have come about as a function of repeated exposure to the experimental arrangements e.g., practice effects– accuracy on task occurs as a function of repeated exposure to the task before the IV is introduced
Testing: internal validity threat
Changes may reflect modifications in the measurement systems rather than effects of the independent variable
E.g., Subjective judgment of human observers, poor integrity of treatment delivery, damaged or no equipment, poor calibration of measurement devices
Instrumentation: internal validity threat
Introduction of the independent variable may coincide with other events in the persons life. Those other events could have produced via fax. E.g., psychiatrist introduce his medication change just as you introduce the behavioral intervention
History: internal validity threat
Natural developmental events or learning experiences may coincide with the introduction of the independent variable to produce the change. E.g., growing older, stronger, healthier, etc.
Maturation right
The assignment of subjects to groups may have biased the outcome even in the absence of any intervention
E.g., self selection bias – participants who for various reasons are more prone to show greater improvement may also be more likely to participate in a study
Perhaps less relevant to SCD because participants serve as your own controls. All controls or individual is exposed to both BL and intervention
Selection bias: internal validity threat
The loss of subjects over time, especially if systematic, May influence the effects
E.g., subjects that tender towards the extreme ends of the measure may leave selectively. Thereby skewing the sample at post test
Perhaps less relevant to SCD because participate serve as their own controls
Attrition : internal validity threat
Continuous assessment – collecting data on the dependent measure for an extended period of time
In contrast to very small samples of the dependent variable
Helps to rule out that changes came about as a function of factors that could have all day performance overtime e.g., testing regression maturation
Measurement: minimizing validity threats
Stabley seeing stability of the target behavior
If levels of the dependent variable remains relatively stable over time before the independent variable is introduce… The likelihood that the change can be attributed to the independent variable increases
Minimizing validity threats:
STABILITY
Immediate effects of the independent variable
The more immediate affects the stronger the case that the IV produced it
Slow effects that appears long after the introduction of the IV call into question whether the change was imminent despite the intervention i.e., intervening events may have caused the change
Immediacy: minimizing validity threats
Demonstration using multiple cases
if the IV affects many subjects in the same manner
A stronger case can be made that it produced the change
Replication: minimizing validity threats
Used for setting target outcomes, e.g., reduction or increase relative to baseline
Sometimes shows that intervention is not needed or maybe unlikely to produce gains
Setting targets: baseline function
Assessment of the dependent variable prior to the introduction or change of the independent variable
Does not necessarily imply the absence of treatment
The longer the baseline the grade of the predictive power
Nothing is gained by unduly long baselines
As long as necessary as short as possible. Generally less should be dictated by qualitative features of the data path… Phase change logic
Movement in the analysis from one level or kind of independent variable to the next level kind of independent variable
Phase changes move the analysis towards an AB design
Predictive
Predict the future level of the target behavior in the absence of the IV or if the IV has no effect
Serves as a criterion to evaluate whether the intervention produces change
Other purposes/benefits of baseline data:
Sometimes simply observing during baseline provides insight into relevant environmental events that can promote treatment development.
E.g., baseline sessions conducted a crossed the day but you notice problem behavior is more likely during morning rather than afternoon baseline sessions
Baseline logic
- Measurement (attrition, continuous assessment, testing, regression, maturation)
2, Stability
- Replication
- Immediacy – immediacy of independent variable
Minimizing validity threats
Extent to which the experiment shows the changes in the behavior are due to the independent variable and not the result of uncontrolled or confounding variables
Examples of confounding variables:
- Starting medication
- Change in home life (divorce, death, new baby, move, etc)
- Change in school/therapy life (new therapists, clinicians, new building, new school, etc
Internal Validity
The degree to which a study’s findings have generality to other subjects, settings, and/or behaviors
-Will this study prove effective if a different population of participants is used?
-Will this study be effective if used with a different type of behaviors?
If the study was done in a clinic, will it be effective if conducted in a school classroom setting? What about in a home environment?
External Validity
In non-concurrent multiple baselines, the baselines are staggered by…..
The number of data points in the same way
Analysis need not be completed concurrently because the chances that the same extraneous variable, rather than that IV reduced the same result is even more remote if the baselines are separated by time
Non concurrent Multiple Baseline
Use of this is inappropriate when the behaviors are interdependent
Multiple baseline
Changing behaviors in one baseline is likely to change behaviors and the other baselines even before the independent variable is presented
Multiple baseline
Most powerful demonstration of functional relationships
Reversal design, A – B – A
• Method of data analysis:
- Visual data analysis.
- sometimes statistical analysis
- Repeated and systematic presentation and removal of a treatment and measurement of behavior while holding other factors constant
- In ABA, valid experimental design and rigorous treatment evaluation are synonymous.
- A variety of designs that use baseline logic to demonstrate the effects of independent variables on the behavior of individual subjects
- !Comparison is made within individuals
- Lends itself to clinical investigation and treatment accountability because participants serve as their own controls.
Single case design
Exposure to the treatment:
Each individual exposed to each level of the IV.
Both baseline AND treatment
Both cases the objective is the same:
To make valid inferences about the effect of an independent variable while, ruling out other possible sources of variability between the experimental condition and the control condition
• Represents the performance of an individual subject
-The one that needs help
Single case design
