Response given antecedent conditional probability equals valid antecedent exposures followed by the defined response within the forward window divided by all valid exposures to that antecedent. Count every exposure, including those without the response. Preserve truncated windows and compare the conditional result with the response's background probability under compatible observation units before drawing a descriptive conclusion.

Write the exposure question first

Write the response given antecedent conditional probability as P(response | antecedent) = valid antecedent exposures followed by the response within the forward window / all valid antecedent exposures. Name the antecedent, response, onset anchor, window, observation settings, and sampling period. This direction asks what happened after the antecedent. It does not answer how often the antecedent occurred before responses unless a separate response-conditioned calculation is built.

Define “noisy transition” in observable terms, including the sound or crowding condition, start and end, setting boundaries, and how simultaneous transitions are coded. Define the response independently. Event labels based on assumed function or intent cannot support a reproducible exposure table.

Enumerate every antecedent exposure

Build the exposure list before looking at response outcomes. Each row should include exposure ID, date and time, setting, source, observer, validity, window-complete status, response inside or outside the window, and relevant access or health context. Counting only antecedents that precede the response would inflate the numerator and erase valid nonresponse exposures.

Reconcile planned observation time, actual observation, valid units, antecedent exposures, complete windows, missing windows, and exclusions. Report whether exposure opportunities were sampled across representative routines. Fifteen exposures from one unusually crowded hour cannot automatically describe other transitions.

Set the forward window and recurrence rule

Choose when the clock starts and how soon the response must begin. For a 20-second onset-to-onset rule, a response beginning at second 21 falls outside the conditional numerator even if it may still be contextually relevant. Preserve exact latency when available so another reviewer can apply the rule.

Repeated or overlapping antecedents need a declared method. A second noisy event might reset the clock, open a second exposure, or remain part of the same transition episode. Simultaneous coded antecedents may create multiple marginal exposures or one compound category. Window selection affects results, as illustrated in contingency space research.

Recalculate Priya’s example

Priya has 15 valid noisy-transition exposures with complete 20-second windows. The response begins within the window after 6. Therefore, P(response | noisy transition) = 6 / 15 = 0.40, or 40%. The other nine valid exposures stay in the denominator and have no qualifying response inside the window.

Across 50 valid background observation units, the response occurs in 9: 9 / 50 = 0.18, or 18%. The observed probability-point difference is 40% − 18% = 22 percentage points. If the units and definitions are comparable, the ratio is 0.40 / 0.18 = about 2.22. Keep 6 of 15 and 9 of 50 beside any comparison because each new event has visible influence.

The background result does not supply the complement cell needed for every causal claim. It shows response presence in all sampled units, including those with and without noisy transitions. A full contingency-space analysis may require response occurrence and nonoccurrence under both antecedent-present and antecedent-absent opportunities.

Treat zero and missing exposures correctly

With fifteen valid exposures and no responses inside the window, the conditional probability would be 0 / 15 = 0%. With no observed noisy transitions, the denominator would be zero and the probability would be undefined. A report should say “No eligible exposure occurred in the sample,” then consider whether observation coverage was adequate.

An exposure at observation end may have a truncated forward window. Observer interruption, obscured video, unavailable timing, or uncertain event boundaries create missing sequence evidence, not a nonresponse. Report complete windows divided by recorded exposures and examine whether missingness clusters in the busiest or most difficult transitions.

Display and describe the pattern

Plot a two-cell exposure bar or dot with 6 of 15 qualifying responses and 9 of 15 without a qualifying response. Add the 18% background result in a separate, clearly labeled mark. A session-level table should show whether all six sequences came from one day or were distributed across routines.

A cautious note might say: “The response began within 20 seconds after 6 of 15 valid noisy-transition exposures (40%). It occurred in 9 of 50 valid background units (18%). The conditional value was 22 probability points higher in this sample. Event definitions, the 20-second rule, clustered settings, and small cell counts limit interpretation.”

Review correlated conditions and causal limits

The sampled difference supports a question about transitions; it does not show that noise caused the response or establish behavioral function. Sound intensity, crowding, unfamiliar people, communication access, pain, fatigue, prior activity, demand changes, and observer coverage may covary with the coded antecedent. The response may also occur after different events.

Antecedent-versus-consequent research demonstrates the value of comparing sequence and background occurrence. Comparative descriptive-method research documents limited correspondence with experimental findings, while functional-assessment ambiguity research distinguishes association from experimental demonstration.

Include accessible client review

Ask Priya, through her preferred communication, about noise, crowding, transition predictability, pain, comfort, sensory needs, and what support would be useful. Keep AAC available under ASHA guidance. Record assent, dissent, and client priorities with their source.

Do not provoke or prolong a difficult transition to fill a denominator. Maintain food, water, bathroom use, mobility, prescribed health care, rest, relationships, and emergency help. Medical and safety concerns route promptly to qualified professionals. The BACB ethics hub and CASP public summary provide professional context.

Clinician checklist and practical limitations

Before interpreting the result, confirm:

  • antecedent and response definitions, onset anchor, window, episode, and overlap rules are fixed;
  • all 15 exposures reconcile to response, nonresponse, truncated, missing, or excluded states;
  • 6 / 15 = 40% and 9 / 50 = 18% reproduce from source records;
  • zero numerators and zero exposure denominators have correct labels;
  • background and conditional observation units are compatible for the stated comparison;
  • displays show raw cells, sessions, coverage, missingness, and correlated context; and
  • accessible client input, noncausal limits, next assessment step, owner, and date are recorded.

Priya’s sample is small and may overrepresent particular transitions. The result changes to 33.3% at 5 of 15 and 46.7% at 7 of 15. A different sequence window, event definition, or setting mix can alter the estimate. Reopen the calculation after source, observer, software, access, health, or protocol changes and retain previous versions.

Related resources

Sources