Comparative, Component, and Parametric Analyses

Three Analyses, Three Questions

Once an analyst knows a treatment works, the next questions are about refinement, and three specialized analyses answer them. These are not separate designs to memorize so much as three purposes layered onto the single-case structures you already know.

A comparative analysis evaluates the relative effects of two or more different conditions. In practice this might mean comparing noncontingent reinforcement (NCR) with differential reinforcement of alternative behavior (DRA), or comparing two prompting strategies. Multielement (alternating-treatments) and reversal formats are common vehicles, but the design must still control for carryover and ensure the conditions are discriminable.

A component analysis identifies which parts of a treatment package are necessary and which are inert. If a package combines antecedent prompts, response blocking, and reinforcement, the analyst can add components one at a time (additive) or remove them one at a time (subtractive) and watch what happens to behavior. The aim is not novelty; it is the simplest effective package — fewer parts mean easier, more faithful implementation.

A parametric analysis evaluates different values of the same variable. Examples include schedule-thinning steps, token exchange ratios, prompt delays, session durations, or reinforcer magnitudes. The question is not whether reinforcement works, but which parameter value produces acceptable outcomes.

A handy mnemonic separates them cleanly. Comparative = different things (apples vs. oranges). Component = parts of one thing (which slices of the apple matter). Parametric = more or less of one thing (how big a slice). If you can restate the analyst's question in one of those three forms, you have identified the analysis — and the rest of the item usually falls into place.

Telling Them Apart

The distinctions are clean once you focus on what is being varied.

A quick exam decision rule:

• If the options differ by kind (two different procedures) → comparative.
• If the options remove or add parts of a single package → component.
• If the options change the amount or value of one variable → parametric.

These are not exotic standalone designs; they are questions asked through the single-case designs you already know. A component analysis often uses a reversal or multielement structure; a parametric analysis often uses a multielement or sequential-phase structure. The labels describe the purpose of the comparison, while reversal, multiple baseline, multielement, and changing criterion describe the structure that supplies prediction, verification, and replication.

Additive vs. Subtractive Component Analyses

Component analyses come in two directions, and the exam may name either.

• In an additive (constructive) component analysis, the analyst starts with a single component and adds pieces one at a time, watching which additions improve behavior. This shows the smallest set needed to reach the goal.
• In a subtractive (dismantling) component analysis, the analyst starts with the full package and removes pieces one at a time, watching which removals cost effectiveness. A component whose removal does not hurt behavior is unnecessary.

Either approach typically rides inside a reversal or multielement structure so each addition or removal is a comparison with verification. A common finding is that a multicomponent plan can be trimmed to its active ingredient — say, the reinforcement contingency — while inert add-ons (extra prompts, a timer) are dropped. The simpler plan is then easier to run with high integrity, which is itself a clinical win, not just a tidiness preference.

Why These Analyses Matter Clinically

Refinement analyses directly serve efficiency and social validity. A treatment package may be effective but too effortful, intrusive, or time-consuming for caregivers to run faithfully in the natural environment. Each analysis offers a remedy:

• A component analysis can strip out an unnecessary piece — for instance, showing that reinforcement alone matches the full package, so response blocking can be dropped — yielding a leaner, more acceptable plan with better procedural integrity.
• A parametric analysis can find the leanest dose that still maintains behavior — for example, identifying that an FR 3 schedule sustains responding nearly as well as FR 1, allowing thinner reinforcement that is easier to deliver and closer to natural contingencies.
• A comparative analysis can select the intervention that is not only effective but also more efficient, less restrictive, or more acceptable to stakeholders.

Throughout, single-case logic still governs. Repeated measurement, stable baselines, planned comparisons, and control of multiple-treatment interference remain mandatory — otherwise the 'refinement' rests on noise.

On the exam, expect a stem describing an analyst who already has an effective treatment and now wants to simplify it, tune it, or pick between options; map the verb (compare / remove-add / change the value) to the matching analysis, and confirm the underlying design still provides verification and replication. A frequent distractor swaps the labels — calling a 'remove a component' study 'parametric,' or calling an 'FR 1 vs. FR 5' study 'comparative.' Anchor on what is varied: the kind of treatment, the parts of one package, or the value of one variable.