Comparative analysis

The question is relative performance

A comparative analysis asks which of two or more defined options performs better under planned conditions. The options might be teaching packages, prompting procedures, schedules, communication-partner responses, or supports.

The analysis can examine several dimensions:

• level and trend of the target response
• speed to a useful effect or mastery
• errors, prompts, and procedural integrity
• generalization and maintenance
• unwanted effects and safety events
• client preference, assent, and burden
• staff time, training, and resource demands

A single outcome rarely settles the choice. One procedure may teach faster while another produces fewer errors or stronger preference. State how dimensions will be weighed.

Match the conditions closely

Differences should mainly reflect the procedures being compared. Keep response definition, materials, opportunity type, session duration, implementer training, setting, data system, and consequence availability as comparable as the question allows.

When conditions must differ, record the difference. A longer session creates more response opportunities. A familiar implementer may change performance. Reporting raw counts without the opportunity denominator can favor the condition with more chances.

Use procedural-integrity measures for each condition. An apparent treatment difference may actually be an implementation difference.

Select a design that fits the procedures

A multielement or alternating-treatment design can compare conditions rapidly when carryover and interaction risks are manageable. A reversal design may support a slower comparison when effects can change with condition withdrawal. A multiple-baseline or adapted sequence can be useful when teaching is durable.

Randomized or counterbalanced condition order can reduce predictable order bias. Distinct cues may help participants and implementers discriminate conditions. Stability, exposure, and stopping rules should be planned before the analysis.

The BACB BCBA Test Content Outline includes rationales for comparative, component, and parametric analyses and covers single-case interpretation. It identifies examination content rather than prescribing one design.

Comparative and component analyses answer different questions

A comparative analysis evaluates whole options, such as Package A versus Package B. A component analysis tests which part of a package contributes to the effect, such as Package A with and without video modeling.

Comparing two packages cannot identify the active ingredient when they differ in several ways. Likewise, finding that one component matters inside Package A does not prove Package A outperforms every alternative.

Define the question narrowly enough that the design can answer it.

A fictional communication-program comparison

Theo chooses a goal of independently requesting help through speech, AAC, or gesture. The team compares two teaching packages during matched activities. Package A uses a model and five-second prompt delay. Package B uses a visual cue and ten-second delay. Both honor recognizable requests and preserve free communication access.

Across twelve eligible opportunities in each condition, Theo requests independently in 7 of 12 Package A opportunities and 10 of 12 Package B opportunities. Procedural integrity is 11 of 12 for A and 12 of 12 for B. Theo selects B for four of six later choice sessions.

Those data favor B in this small comparison. They do not establish equivalence, identify the active component, or predict performance in another setting. A changed delay and cue both differ, and implementer fidelity also differs. The team gathers maintenance and generalization data before revising the plan.

Interpret magnitude, consistency, and overlap

Visual analysis considers level, trend, variability, immediacy, overlap, and consistency across similar phases. The WWC Single-Case Design Technical Documentation describes internal-validity standards and visual evidence of a relation. Its framework is research-review guidance, not a clinical scoring rule.

Also report raw counts, opportunity definitions, prompts, exclusions, fidelity, adverse events, and person-reported experience. A percentage can conceal three opportunities in one condition and thirty in another.

Avoid declaring two procedures “the same” because their averages look close. Equivalence requires a design and analysis built for an equivalence question. A weak or short comparison may simply lack resolution.

Use a prospective decision rule

Before starting, define:

1. eligible opportunities and minimum exposure
2. primary and secondary outcomes
3. safety and assent stop rules
4. acceptable fidelity
5. meaningful difference or decision threshold
6. follow-up needed for maintenance and generalization

If the data remain mixed, continue only when clinically appropriate, refine measurement, or choose based on preference, burden, and contextual fit. “Inconclusive” is a valid result.

Archive the original plan, raw data, fidelity records, and decision memo. A later reviewer should be able to reconstruct which conditions were compared and why the team selected the next step.