Inconclusive ABC probability results should preserve the conflicting session or context patterns, raw cells, small denominators, missing conditions, and measurement limitations. State which relations were examined and why no stable descriptive conclusion is supported. Avoid forcing a function label from pooled percentages. Name the next safe assessment, observation, medical, access, or interdisciplinary step and its owner.

Define what was compared

An inconclusive report still needs a reproducible method. Name the response, event, conditional direction, anchor, sequence window, background unit, sessions, settings, observers, and coverage. Preserve the question that led to the analysis, such as whether attention occurred more often inside a response-following window than across ordinary observation intervals.

Predeclare which contexts will remain separate. Pooling across weeks, settings, or protocol versions can average conflicting patterns into an apparently stable middle. Stratification decided only after seeing results can also create misleading chance findings. Record the original plan and any exploratory follow-up distinctly.

Reconcile every evidence cell

For each week or context, report response windows with and without attention, background units with and without attention, truncated windows, missing observation, and excluded cells. Keep planned and valid coverage nearby. A percentage without these counts cannot show whether the pattern rests on two exposures or twenty.

Verify that definitions, timing, interval length, and observation opportunities remained compatible. If they changed, report separate protocol versions and avoid a numerical comparison that implies sameness. Preserve source corrections and the table version used in the clinical meeting.

Verify Hollis’s week-level arithmetic

In week one, attention follows 4 of 6 valid responses: 4 / 6 = 66.7%. Background attention occurs in 10 of 30 intervals: 10 / 30 = 33.3%. The conditional-minus-background difference is 33.3 percentage points, and the descriptive ratio is 2.0.

In week two, attention follows 2 of 5 responses: 2 / 5 = 40%. Background attention occurs in 12 of 25 intervals: 12 / 25 = 48%. The difference is −8 percentage points, and the ratio is about 0.83. The direction reverses: week one conditional probability is higher than background, while week two is lower.

Pooled conditional probability is (4 + 2) / (6 + 5) = 6 / 11 = 54.5%. Pooled background probability is (10 + 12) / (30 + 25) = 22 / 55 = 40%. The pooled difference is 14.5 points. That aggregate hides the week-level reversal and should remain secondary.

Name why the result is inconclusive

Several explanations can remain open: small response denominators, different settings, uneven observation schedules, changing attention availability, definition drift, missing health or access context, observer differences, or genuine variation across weeks. State which have evidence and which are hypotheses.

Avoid treating “inconclusive” as a data-quality failure or as proof that no relation exists. It means the sampled descriptive evidence does not support one stable interpretation under the declared protocol. The conclusion can be clinically useful because it prevents an unsupported function label.

Handle zeros, gaps, and incompatible cells

A week with no response exposures has an undefined conditional probability, not 0%. A week with responses and zero attention sequences has a calculable 0%. Truncated follow-up windows remain outside the valid response denominator. Missing background intervals remain outside the background denominator and visible in coverage.

When event categories overlap, preserve marginal or exclusive rules. Do not recode “attention plus break” differently between weeks to make tables align. If timestamp precision changes, avoid claiming that the same sequence window was measured.

Graph conflict instead of smoothing it away

Use paired week-level dots or bars for conditional and background probabilities, with raw fractions printed next to each mark. Connect neither series if the weeks differ materially in context. Add pooled values as lighter secondary references, never as the only display.

Suggested wording: “Week one showed attention after 4 of 6 responses (66.7%) versus 10 of 30 background intervals (33.3%). Week two showed 2 of 5 (40%) versus 12 of 25 (48%). Because the direction reversed and response denominators were small, no consistent sampled relation is supported. Pooled values are reported only for completeness.”

Keep function and causation unresolved

Inconclusive ABC probability results do not support an automatic-reinforcement label, an attention function, or absence of function. Descriptive associations can reflect correlated events, setting differences, pain, communication access, staffing, task conditions, and sampling. Comparative descriptive-method research reports limited correspondence with experimental findings.

Predictor research, contingency-space analysis, and lag-sequential work support explicit background, timing, and cell comparisons. They do not create a universal rule for resolving conflict.

Select a safe next evidence step

Route pain, health change, adverse effects, or immediate safety concerns promptly to appropriate professionals. For measurement problems, repair definitions, clocks, observer training, or coverage and collect new independent evidence. For context gaps, expand the prospective sampling frame with Hollis’s agreement and access needs.

A qualified clinician may consider direct client input, record review, interdisciplinary consultation, or other assessment within competence and authorization. Observation cannot require recreating danger, delaying help, or withholding a meaningful communication response merely to obtain a clearer cell.

Review access and meaning with Hollis

Ask Hollis through a preferred communication method what attention, response events, settings, discomfort, and supports mean in daily life. Keep AAC continuously available under ASHA guidance. Record assent, dissent, priorities, and corrections.

Maintain food, water, bathroom use, mobility, prescribed health care, rest, relationships, and emergency help. The BACB ethics hub and CASP summary offer professional context. Medical, safety, privacy, payer, and legal matters remain with qualified owners.

Clinician checklist and practical limits

Before releasing the inconclusive report, confirm:

  • response, event, direction, window, background unit, weeks, and contexts are defined;
  • conditional, background, missing, excluded, and truncated cells reconcile by week;
  • week-one 66.7% versus 33.3% and week-two 40% versus 48% reproduce;
  • pooled 6/11 = 54.5% and 22/55 = 40% remain secondary to the reversal;
  • graphs show raw fractions, small denominators, coverage, and contextual differences;
  • wording states no consistent sampled relation without assigning a function; and
  • accessible client input, safe next step, qualified owner, and review date are recorded.

Hollis’s response denominators of six and five are highly sensitive to one sequence. Unknown context and missing settings may change the interpretation. Additional descriptive data can remain inconclusive. Reopen the report when source records, protocols, access, health, settings, observers, or software change, preserving every earlier conclusion and its evidence.

Related resources

Sources