Measurement and Data Collection Flashcards

A researcher is measuring the duration of stereotypic behavior in a client with ASD. During a 30-minute session, the behavior occurs in 4 separate bouts lasting 3 min, 7 min, 2 min, and 8 min. The client is absent for 5 minutes mid-session. What is the correct percentage of session time the behavior occurred, accounting for the absence?

Answer: 66.7%

Total observed session time = 30 - 5 = 25 minutes. Total behavior duration = 3+7+2+8 = 20 minutes. Duration percentage = (20/30) × 100 = 66.7%. The denominator uses the full scheduled session (30 min), not just observed time, unless the recording method explicitly excludes non-observation periods. Using 25 min would inflate the percentage inappropriately.

An ABAT practitioner is evaluating two observers' data on a continuous recording of vocal behavior. Observer A recorded 42 intervals with behavior; Observer B recorded 38. Their agreements (both scored behavior in the same interval) totaled 35, and their disagreements (one scored behavior, the other did not) totaled 12. Which IOA formula should be used, and what is the resulting value?

Answer: Occurrence IOA; 83.3%

When a behavior has a low base rate or the researchers are specifically concerned with occurrence reliability (not non-occurrence), Occurrence IOA is most appropriate. Occurrence IOA = Agreements on occurrence / (Agreements on occurrence + Disagreements) = 35 / (35+7) — but here disagreements split: 5 instances where A said yes/B no, and 7 where B said yes/A no. Occurrence IOA = 35/(35+12) = 35/42 = 83.3%. Total count IOA (38/42=90.5%) inflates agreement by ignoring interval alignment.

A behavior analyst is designing a measurement system for a behavior that occurs at very high rates (estimated 80–120 times per hour) and lasts only 1–3 seconds each instance. Which measurement system is MOST appropriate, and what is the primary reason?

Answer: Partial-interval recording, because it is less labor-intensive and captures high-rate brief behaviors without requiring precise onset/offset timing

For very high-rate, brief behaviors, continuous frequency recording is practically difficult (observer fatigue, counting errors) and time-sampling methods like MTS will grossly underestimate occurrence. Partial-interval recording is the most defensible compromise: it acknowledges that behavior did or did not occur without requiring precise tallying of each instance. It overestimates prevalence but is acceptable when the goal is detecting trends rather than exact rates. Whole-interval recording would severely underestimate a 1–3 second behavior.

During a functional analysis, a clinician uses 10-second partial-interval recording across four 5-minute conditions. SIB is recorded in 18 of 30 intervals in the attention condition and 6 of 30 in the control. A colleague argues the data overestimate true frequency. Which statement BEST captures the methodological nuance?

Answer: Partial-interval recording overestimates prevalence; the overestimation is greater when behavior duration is shorter relative to the interval, making the attention vs. control contrast directionally valid but not quantitatively precise

Partial-interval recording always overestimates prevalence because any occurrence during the interval—no matter how brief—scores the entire interval as positive. The overestimation is largest when behaviors are brief relative to interval length. While comparing across conditions does attenuate the practical impact, it does not eliminate the quantitative imprecision. The directional contrast (attention > control) can still support functional conclusions, but absolute prevalence estimates are inflated.

A supervisor reviews a trainee's frequency data collected over sessions of varying lengths: 45 min (12 responses), 60 min (14 responses), 30 min (9 responses), and 90 min (19 responses). The trainee reports mean frequency as 13.5 responses per session. What is the correct criticism, and what is the proper metric?

Answer: Mean count is inappropriate across unequal session lengths; rate per minute should be used: approximately 0.26, 0.23, 0.30, and 0.21 responses/min, mean rate ≈ 0.25/min

When session lengths vary substantially (here from 30 to 90 minutes — a 3× range), raw frequency counts are not comparable across sessions. Reporting mean count (13.5) misrepresents the data. The correct metric is rate (responses per unit time). Calculating per minute: 12/45=0.267, 14/60=0.233, 9/30=0.300, 19/90=0.211. Mean ≈ 0.253/min. Rate per hour is also acceptable but per-minute is standard in ABA. The trainee's error is a fundamental measurement validity issue.

A researcher uses a multiple-probe design to assess acquisition of a 5-step task chain. They collect data using task-analytic assessment: each step scored as + (independent) or – (prompted/error). After baseline, the client achieves 5/5 on step 1 during training, but steps 2–5 remain at 0/5. The researcher probes all steps once per phase to conserve resources. A reviewer flags a potential confound. What is the MOST likely confound the reviewer identified?

Answer: Infrequent probing of untrained steps reduces the ability to detect spontaneous generalization or sequence learning, potentially confounding the functional relation demonstration

In a multiple-probe design, probes of untrained steps are taken infrequently (unlike continuous baselines in multiple baseline designs). The confound is that if the client spontaneously generalizes skills or benefits from sequence-learning exposure during training of earlier steps, the infrequent probes may miss this trend—or the eventual improvement during probe phases may be attributed to the intervention when maturation or practice effects are the true cause. This threatens internal validity by making it difficult to rule out alternative explanations for behavior change in later steps.