Experimental Design and Data Analysis
What this domain tests
Experimental design and data analysis tests single-subject methodology: reversal, multiple baseline, alternating treatments, changing criterion, and withdrawal designs; threats to internal validity; and visual analysis of level, trend, variability, immediacy, and overlap.
Core rules
- Baseline logic requires a stable baseline before intervention; a trended baseline confounds interpretation.
- Reversal (ABAB) demonstrates experimental control by showing the behavior returns when treatment is withdrawn, but it is inappropriate for dangerous or irreversible behavior.
- Multiple baseline across behaviors, settings, or participants staggers the intervention; change should occur only when the intervention is introduced.
- Alternating treatments compares two or more conditions rapidly; counterbalance order to control for sequence effects.
- Changing criterion requires stepwise criteria and documentation that behavior tracks each step.
- Visual analysis: level (mean), trend (slope), variability, immediacy of change, and overlap.
- Confounds include history, maturation, instrumentation, observer drift, and concurrent events.
Worked example
A team introduces a token economy in three classrooms sequentially. On-task behavior improves in Classroom A only after introduction, in B only after its introduction two weeks later, and in C only after its introduction. Reasoning: the staggered onset across three independent settings reduces the plausibility of history or maturation; a simple AB design in one room would not.
Traps
Claiming causation from an AB design; reversing a dangerous behavior; using alternating treatments without counterbalancing; interpreting a trended baseline as an intervention effect; ignoring IOA and procedural integrity when judging effects.