Data Collection and Analysis Flashcards

A behavior technician is collecting data on self-injurious behavior using a 10-second partial interval recording method during a 30-minute session. The behavior occurs continuously for 4 full minutes at one point, then in brief 3-second bursts scattered throughout. Which limitation of this data collection method is MOST relevant to this scenario?

Answer: Partial interval recording overestimates the occurrence of behaviors that occur in brief, frequent bursts but underestimates behaviors that occur in long, continuous episodes

Partial interval recording scores an interval as positive if the behavior occurs at any point during it, regardless of duration. This means a 4-minute continuous episode scores the same as a single 3-second burst across multiple intervals — overestimating brief behaviors (every short burst 'fills' an interval) while the 4-minute stretch can only account for a limited number of intervals. This asymmetric distortion is the core methodological limitation in this scenario.

A BCBA reviews a graph of a client's on-task behavior measured via momentary time sampling (MTS) at 5-minute intervals across a 45-minute session. The data shows 70% on-task behavior. The BCBA then reviews video and finds the client was on-task for only 25% of the total session time, but happened to be on-task during most of the observation moments. Which conclusion is MOST accurate?

Answer: MTS provides an estimate of the proportion of time the behavior occurs, but its accuracy depends on whether behavior occurrence is independent of the sampling moments — a systematic pattern can cause substantial over- or underestimation

MTS is designed to estimate the proportion of time a behavior occurs, and it is unbiased only when behavior occurrence is independent of the sampling moments. In this scenario, the client's on-task behavior happened to coincide with most observation moments, causing a large overestimate (70% MTS vs. 25% actual). This is a known limitation — not necessarily reactivity — and illustrates why MTS accuracy depends on the assumption of independence between behavior and sampling schedule.

During a discrete trial training session, a technician is instructed to collect trial-by-trial data using a '+' for independent correct responses and a '-' for errors or prompted responses. At the end of the session, the technician calculates 80% accuracy. However, the BCBA notes that the last 5 trials were all independent correct responses. Which data analysis consideration is MOST important when interpreting this session's results?

Answer: Session-level percentage accuracy masks within-session trends; the terminal performance (last 5 correct) may indicate acquisition momentum that a single aggregate score obscures

An 80% session-level accuracy score is an aggregate that loses information about the sequence of responses. If errors clustered early in the session and correct responses clustered at the end, the terminal behavior pattern suggests the client may be acquiring the skill — a very different clinical picture than if errors were randomly distributed throughout. Examining within-session trends from trial-by-trial data is critical for making accurate programming decisions.

A technician is collecting ABC (Antecedent-Behavior-Consequence) narrative data on aggression. After two weeks, the BCBA attempts to analyze the data but finds it difficult to identify reliable patterns. Which of the following factors MOST likely compromises the utility of ABC narrative data for functional analysis purposes?

Answer: Narrative ABC data is subject to observer bias, incomplete recording, and retrospective interpretation, making it vulnerable to missing low-rate antecedents and inconsistent consequence categorization across observers

Narrative ABC data is an indirect and descriptive method that relies on the observer accurately noticing, remembering, and recording all relevant antecedents and consequences in real time. This is prone to observer bias (recording events consistent with existing hypotheses), incomplete recording (missing events during high-demand periods), and inconsistent categorization of consequences across different technicians. These limitations accumulate over time and make reliable pattern detection difficult, regardless of session count.

A behavior technician is asked to take interobserver agreement (IOA) data on challenging behavior using frequency counts. Observer A records 12 occurrences and Observer B records 15 occurrences during the same session. Using the smaller-to-larger ratio method, the IOA is 80%. The BCBA then requests that the technician also calculate trial-by-trial IOA using the occurrence-nonoccurrence breakdown. Why might trial-by-trial IOA provide more clinically meaningful information in this context?

Answer: The total-count ratio method can mask systematic disagreements — observers may agree on total count while disagreeing on which specific occurrences were counted, and occurrence/nonoccurrence analysis reveals whether agreement is concentrated in occurrence intervals or non-occurrence intervals

Two observers can arrive at similar total counts through entirely different recordings — one might count 12 behaviors at different moments than another who counted 15. The total-count ratio (12/15 = 80%) says nothing about whether they agreed on the same behavioral episodes. Occurrence/nonoccurrence IOA within time blocks or trials reveals whether agreements are real (both scored the same events) or coincidental. It also shows whether disagreements cluster in occurrence or nonoccurrence intervals, which has different clinical implications for measurement reliability.

A technician is graphing a client's data on a Standard Celeration Chart (SCC). The BCBA notes that the data path shows a 'bounce' (variability) that spans two log cycles. Which interpretation of this pattern is MOST consistent with SCC conventions?

Answer: On a Standard Celeration Chart, a two log-cycle bounce means the highest data point is 100 times greater than the lowest data point within that data set, indicating extremely high variability that may compromise interpretation of celeration

The Standard Celeration Chart uses a semi-logarithmic scale where each log cycle represents a 10-fold change. A bounce spanning two log cycles means the ratio between the highest and lowest data points within the set is 10² = 100. For example, if the lowest point is 2 and the highest is 200, that is a two log-cycle bounce. This level of variability is severe and indicates that the celeration line (rate of change over time) may not meaningfully represent the data, and program adjustments are likely needed before trends can be reliably interpreted.