Consequence given response conditional probability equals valid response events followed by the defined consequence within the selected window divided by all valid response events eligible for that window. State the response timestamp, consequence code, window, and treatment of overlaps or truncated observations. Report raw counts, missing sequences, and the consequence's background probability with the result.

Define the sequence precisely

Write the formula as P(consequence | response) = response events followed by the consequence within the declared window / all response events with a valid complete window. Define the response, consequence, sequence anchor, maximum delay, observation settings, and eligible dates before reviewing the result.

Choose response onset, offset, or another observable anchor and apply it consistently. A ten-second window after onset can overlap the response itself; a window after offset cannot. State whether a consequence already underway at the anchor qualifies, whether its onset must fall inside the window, and how repeated response events affect an active window.

Build the response exposure list first

Create one row for every recorded response event before marking consequences. Include response ID, time, setting, source record, observer, complete-window status, consequence codes, missingness, and notes about access or health context. This prevents consequence presence from determining which responses enter the denominator.

Reconcile planned observation, observed time, valid time, response events, complete windows, truncated windows, and excluded events. A response at session end may lack ten seconds of follow-up. Keep it as truncated evidence instead of scoring “no attention.” If the source is missing for part of a window, follow the predeclared validity rule.

Code overlapping and repeated events

Several consequences may occur after one response. With marginal coding, the same response can count toward attention and break, so percentages can sum above 100%. With mutually exclusive coding, define categories such as attention only, break only, both, other, and none. Select the method from the question and retain the raw cells.

Repeated responses create another choice. A new response might start a separate window, reset the existing window, or remain part of one response episode. Define episode boundaries and test observer agreement when the distinction affects the denominator. Window rules from contingency space analysis illustrate how the selected span can change a descriptive result.

Verify Owen’s arithmetic

Owen has 12 response events with complete ten-second follow-up windows. Attention begins inside the window after 7. The consequence given response conditional probability is 7 / 12 = 0.5833, reported as 58.3% to one decimal place. The remaining five complete windows did not contain coded attention; any truncated response windows are reported separately and stay outside the denominator.

Attention also occurs in 20 of 60 valid background intervals: 20 / 60 = 0.3333, or 33.3%. The conditional denominator is 12 response windows, while the background denominator is 60 intervals. These units remain labeled. A probability-point comparison is useful only after confirming that the event definition and observation coverage are compatible.

One additional attention sequence would change the conditional result to 8 / 12 = 66.7%; one fewer would produce 6 / 12 = 50%. That sensitivity makes the small response denominator visible. Preserve 7 of 12 beside 58.3% in every display.

Handle zero cells and missingness

If none of 12 valid response windows contains attention, the result is 0 / 12 = 0%. If no valid response event occurs, the denominator is zero and the probability is undefined. “No observed response exposure” is the accurate result; 0% would falsely claim that valid exposures occurred without attention.

Missing windows can distort the result when they cluster near staff changes, busy routines, certain settings, or longer response episodes. Report complete windows divided by all recorded response events. Avoid treating missing video, obscured timing, or observer absence as “consequence absent.” Recalculate after resolving a source record and retain the prior version.

Graph and write the result

A paired display can show 7 of 12 response windows with attention and 20 of 60 background intervals with attention. Include raw counts, percentages, observation coverage, and the ten-second window in the title or note. A session-level panel can reveal whether seven sequences occurred across many sessions or were concentrated in one.

Suggested wording: “Attention began within ten seconds after 7 of 12 response events with complete follow-up windows (58.3%). Attention occurred in 20 of 60 valid background intervals (33.3%). Five response windows had no coded attention; truncated windows were excluded and reported separately. The sampled association depends on the response anchor, window, and interval system.”

Keep functional interpretation bounded

The observed sequence does not establish that the response produced attention or that attention reinforced the response. Attention may be common throughout the setting, both events may follow another condition, or observation may miss relevant communication, pain, task, social, or health variables. Antecedent-versus-consequent research supports comparing sequence and background occurrence.

Comparisons of descriptive methods document limited correspondence with experimental findings, and research on reducing functional-assessment ambiguity separates observed association from experimental demonstration. The result can guide a qualified next assessment question, not announce behavioral function.

Review access and meaning with the client

Ask Owen through accessible communication what events, discomfort, preferences, or communication attempts the codes may have missed. Keep AAC continuously available under ASHA guidance. Observation must not limit food, water, bathroom use, mobility, prescribed care, rest, relationships, or emergency help.

The BACB ethics hub and CASP public summary offer professional context. Medical, privacy, payer, safety, and legal questions remain with their qualified owners. Record assent, dissent, client priorities, and access changes as evidence, not as obstacles to the calculation.

Clinician checklist and practical limits

Before using the result, confirm:

  • response and attention definitions, anchor, window, episode, and overlap rules are explicit;
  • every response is accounted for as complete, truncated, missing, or excluded;
  • 7 of 12 and 20 of 60 reproduce from source timestamps and interval records;
  • zero numerators and zero denominators retain different meanings;
  • background and conditional units, coverage, settings, and dates are labeled;
  • graphs show raw cells, session distribution, missingness, and small-denominator sensitivity; and
  • the review states its noncausal limit, accessible client input, next step, owner, and date.

The 58.3% result belongs to Owen’s sampled observations and ten-second rule. A different anchor or window can change it, and twelve responses provide limited precision. Descriptive data cannot confirm function or forecast another setting. Reopen the worksheet when definitions, source data, timing, access, health, observer, or software changes, and preserve earlier calculations for review.

Related resources

Sources