Analysis of Behavior ch 3

3 kids of replication in single subject experiments

1. evaluate replication within individuals. 2. replication important in behavior analysis is replication across individuals 3. replication across labs

four defining features of single-subject experimental designs

1. focus is on the behavior of the individual, the single subject 2. every individual receives the intervention 3. behavior is repeatedly measured in each phase until we can confidently predict what will happen if the independent variable is turned ON or OFF 4. single-subject experimental designs evaluate internal validity by evaluating if the independent variable produces systematic and replicable effects on behavior.

Four Types of Single-Subject Experimental Designs

Comparison (A-B) Design Reversal (A-B-A) Design Alternating-Treatments Designs Multiple-Baseline Designs

two key advantages of visual analysis over inferential statistics

First, it requires that the behavior analyst develop methods for producing large changes in behavior; that is, changes that are apparent to the naked eye. A second advantage of the visual analysis is that it recognizes the inherent subjectivity of the scientific enterprise

weaknesses of the group experimental design

First, when the independent variable is a therapeutic intervention, no one wants to be assigned to the control group The second weakness of group experimental designs is that focusing on the behavior of the group means we are not studying the behavior of the individual. The third weakness of group designs is that the behavior of the experimental and control groups differ simply because the people (or nonhuman animals) assigned to the two groups are different. The final weakness of group experimental designs is their reliance on inferential statistics to evaluate if the independent variable changed behavior.

three simple steps to guide the visual analysis. These should be viewed as guidelines, not as rules that must be rigidly followed.

Step 1: Draw a trend arrow through the baseline data to predict what will happen if the independent variable is never turned ON. Step 2: Evaluate if behavior in baseline is too variable (bouncy) to have confidence in the prediction of the trend arrow. Step 3: Draw trend or level lines through the intervention data. Evaluate if there is a convincing change in trend or level (whichever change is of interest).

Simulation Modeling Analysis

The SMA was designed for use with single- 3 subject time-series data and can be conducted with freely available software. The SMA tests for changes in trend and level, making it an ideal supplement for the visual analysis. Best of all, when behavior analysts supplement their visual analysis with results of the SMA, or another inferential statistical analysis, social scientists outside of behavior analysis will take the findings more seriously.

how to combat the weakness that the behavior of the experimental and control groups differ simply because the people (or nonhuman animals) assigned to the two groups are different.

The way that group designs overcome this is to randomly assign subjects to the experimental and control groups. The logic, based in probability theory, is that random assignment will equalize the two groups on variables that might affect behavior. Researchers need large groups (>30) subjects (lots of coin flips) to overcome all this between-groups noise

trend

a systematic change in behavior over time

comparison design

arranges a baseline (A) phase, followed by an experimental (B) phase. This simple OFF-then-ON design is sometimes referred to as an A-B design quasi-experimental design; "quasi" because a well designed experiment will rule out confounds

In a single-subject experimental design, it is always true that every individual will experience the _________________ and experimental (intervention) phases.

baseline

systematically

behavior always increased when the intervention began

In the ___________________ design, behavior is compared between a baseline and a single intervention phase.

comparison

If another variable changed when the independent variable was manipulated, that other variable could explain why behavior changed. That other variable if referred to as a __________________.

confound

If this intervention was designed to decrease the frequency of a problem behavior, then we might be tempted to conclude that it worked. However, something else might be responsible for this change in trend and level. These "something else" variables are referred to as _________________.

confound

Comparison designs are often referred to as quasi-experimental designs because they do not rule out _____________ by repeatedly demonstrating that the independent variable has a systematic effect on behavior.

confounds

multiple baseline design

evaluates the functional relation between an independent variable and behavior by conducting a series of time staggered A-B comparisons either across behaviors, across situations, or across individuals.

error bars

extend above each of these averages represents the variability in the data

single subject designs

focus on the individual and most commonly used by behavior analysts expose individuals to baseline (independent variable OFF) and experimental (independent variable ON) phases to determine if the independent variable systematically and reliably changes behavior.

group experimental designs

focuses on the behavior of the group and most often used in social and behavioral sciences they evaluate if the behavior of a treatment group (independent variable ON) is statistically significantly different from that of a control group (independent variable OFF). If so, then the difference is attributed to the independent variable.

In ________________ experimental designs, the focus is on the behavior of the group, not the individual.

group

In the same graph, if the independent variable is not manipulated, our best prediction is that this ___________ trend will continue into the intervention phase.

increasing

Good experiments are those that can rule out confounds. These experiments have high ______________ ______________.

internal validity

These data come from a comparison (A-B) design. This design does not allow us to rule out confounds. Therefore, we must conclude that the experiment has low _____________ _____________.

internal validity

When an experiment demonstrates that behavior changed because the independent variable was turned ON and OFF, that experiment has high ________________ ________________.

internal validity

visual analysis

involves looking at graphs of time- series single-subject behavior to evaluate if a convincing change occurred when the independent variable was introduced

problems with statistics

know the difference between clinically significant and statistically significant core of the replication crises very subjective

When confounds cannot be ruled out, the experiment has ____________ internal validity.

low

In a _______________-______________ design, the effects of the independent variable may be replicated across behaviors, across situations, or across individuals.

multiple baseline

The _________________-____________________ design is used either when it would be unethical to turn OFF the independent variable or when the independent variable is anticipated to produce a lasting (irreversible) effect.

multiple baseline

Behavior is stable

over repeated observations, there is little "bounce" and no systematic trend

independent variable

publicly observable change, controlled by the experimenter, which is anticipated to influence behavior in a specific way

In a single-subject experimental design, internal validity is assessed through _____________________. For example, if the independent variable systematically influences behavior every time it is turned ON and OFF, then the experiment has high internal validity.

replication

Two of the single-subject experimental designs establish internal validity by turning ON and then OFF the independent variable. The first of these is the ____________ design. The other is the _______________________-______________________ design.

reversal; alternating-treatments

In a _____________-____________ experimental design, the focus is on the behavior of the individual, not the group.

single subject

In group experimental designs, inferential _______________ are used to decide if behavior changed when the independent variable was manipulated.

statistics

systemically

that is, behavior always increased when the intervention began

alternating-treatments design,

the independent variable(s) is turned ON and OFF rapidly to evaluate if this systematically and repeatedly changes behavior

Reversal (ABA) design

the individual's behavior is evaluated in repeatedly alternating baseline (A) and experimental (B) phases. This design is also known as an A-B-A design.

level

the prevalence of the behavior during the stable portion of a phase/condition

multiple-baseline across-behaviors design

time-staggered A-B replications are conducted across behaviors

multiple-baseline across-situations design

time-staggered A-B replications are demonstrated across situations

multiple-baseline across-participants

time-staggered AB replications are demonstrated across participants

In the graph to the right, there is a systematic increasing _______________ in behavior during the baseline phase.

trend

two ways in which behavior can change

trend and level

confounds

variables that are influencing behavior, but are not controlled by the researcher

internal validity

when an experiment provides clear evidence that a functional relation exists between Successful behavioral experiments allow us (and a jury of skeptical scientists) to conclude that there is a functional relation between the independent variable and behavior. When the independent variable is a therapeutic intervention, a successful experiment allows us to conclude that the intervention (and not something else) positively influenced behavior.

internal validity

when an experiment provides clear evidence that a functional relation exists between the independent variable and behavior change