Can Families Ask How ABA Data Are Averaged?

An average needs a defined question

Before interpreting an average, ask what it summarizes. It may describe responses per session, duration per hour, opportunities across a week, caregiver ratings, or values from several settings. The unit and time window determine what the number can mean.

The CASP public summary places assessment, treatment planning, implementation, and evaluation within individualized ABA treatment for people diagnosed with autism. It does not prescribe one averaging method.

Weighting changes the result

Suppose one session has 1 success in 2 opportunities and another has 8 in 8. Averaging the session percentages gives 75%: (50% + 100%) / 2. Pooling the opportunities gives 90%: 9 / 10. The first gives each session equal weight. The second gives each opportunity equal weight.

Ask which weighting matches the question. Keep the two session counts available so a polished percentage does not conceal unequal exposure.

Missing values and outliers stay visible

A canceled session, a session with no eligible opportunity, and an unrecorded session should have different labels. An average over only completed records can look stronger when difficult or inaccessible sessions are missing.

Also inspect the range and individual points. One unusually long event can dominate a duration mean. A median, range, raw series, or setting-specific view may answer a different question.

Clinical interpretation belongs to a qualified role

The BCBA Test Content Outline covers measurement, graphing, visual analysis, validity, and data-based decisions as examination content. It is not an averaging rule or authority to practice.

The BACB Ethics Code addresses documentation, data evaluation, client involvement, and understandable communication for covered behavior analysts. Ask the qualified clinician why the summary supports the stated decision and what information it leaves out.

Reproduce the average before interpreting it

A family should be able to follow the calculation without reverse-engineering the graph. Start with a small table that has one row per observation and columns for the date, setting, numerator, denominator, unit, support level, and exclusion status. Add the formula used for the displayed value. Then calculate it independently.

For example, Noor has three school-practice observations: 3 of 4, 2 of 8, and 4 of 4 opportunities. The session percentages are 75%, 25%, and 100%. Their unweighted mean is 66.7%. Pooling all eligible opportunities produces 9 of 16, or 56.3%. Both calculations are arithmetically sound. They emphasize different things.

The unweighted mean describes the average session percentage and gives the four-opportunity and eight-opportunity sessions equal influence. The pooled percentage describes performance across all recorded opportunities and gives the longer session more influence. A report should name the choice and explain why it serves the clinical question.

Match the summary to the pattern

The arithmetic mean is familiar, but it can be sensitive to a very high or low observation. The median may better describe a typical session when a few extreme values are present. A rate can account for unequal observation time. A percentage can account for unequal opportunity counts when both parts are disclosed. A cumulative total may answer a workload question while hiding recent change.

Families do not need to select the statistical method themselves. They can ask the clinician to show how the conclusion changes under a reasonable alternative. If the decision reverses when one observation is removed or when pooled and unweighted values are compared, that sensitivity is important evidence.

Keep phase and context boundaries intact

Avoid averaging together baseline, teaching, maintenance, and generalization observations unless the combined value answers a clearly stated question. The same caution applies to clinic, home, school, telehealth, and community settings. A single overall percentage can conceal that a skill is strong with one partner and rarely observed with another.

Before accepting a summary, check:

• the exact observation window and phase
• whether every included row used the same definition
• which sessions had no eligible opportunity
• whether prompts and ordinary supports changed
• whether canceled, missing, or invalid records were excluded
• whether a correction changed the underlying values
• whether the person and family recognize the pattern in daily life

The result should remain connected to the raw observations. An average is a compact description, not a substitute for the record that produced it.

If the displayed result cannot be reproduced

Begin with the narrowest possible discrepancy. Identify the graph, date range, measure, displayed average, and values you used. Ask whether the system applies rounding, weighting, exclusions, carry-forward rules, or a different time zone. A transparent answer should point to a documented rule rather than a guess.

If the values are incorrect or the rule was applied inconsistently, the practice should preserve the earlier output, correct the source or calculation through its authorized process, regenerate affected reports, and decide whether any clinical conclusion needs review. The person who repairs a formula does not automatically decide what the changed graph means.

Suppose a quarterly report shows 82%, while the disclosed weekly values produce 76%. The difference may come from pooling all opportunities rather than averaging the four weekly percentages. If that is the intended method, the report can label it. If no disclosed method produces 82%, the result needs reconciliation before it supports a decision.

Agree on the summary before the next review

Families can ask the team to add a short measurement note to the plan or progress-report template. It can state the primary unit, aggregation method, observation window, missing-data treatment, setting breakdown, and rounding rule. The note reduces future confusion and keeps different staff from applying different calculations.

At the next meeting, ask the clinician to show one live example from raw values to final graph. Confirm which pattern would lead to continued observation, a plan review, or another type of assessment. The strongest outcome is not agreement on a favorite number. It is a shared understanding of what the summary measures, what it omits, and how it informs the person's care.

Keep the example with the meeting notes. At the following review, confirm that the same calculation still applies and that any new phase, setting, or support is labeled separately. This small record gives families and clinicians a stable reference when staff, software, or reporting periods change.

It also makes future corrections easier to explain and validate against the original evidence.

Questions families can use

Ask for the numerator and denominator, the raw session values, the unit, the dates, the inclusion rule, missing-data reasons, weighting, range, setting breakdown, phase changes, and the person responsible for interpretation. A useful report lets a reader reproduce the summary from the disclosed values.