- Applied Behavior Analysis Measurement and Data Collection Questions and Answers Flashcards

A behavior analyst is measuring stereotypic hand-flapping using a 10-second partial interval recording system across a 30-minute session. The behavior occurs for 3 continuous seconds at the start of each minute but never recurs within the same minute. What will the recorded prevalence be, and how does it compare to the true prevalence?

Answer: Recorded: 100%; True: ~16.7% — partial interval systematically overestimates

In partial interval recording, an interval is scored if the behavior occurs at ANY point within it. Since the behavior occurs once every minute (every 6 intervals), each of those 6-interval minutes has the first interval scored. But the behavior only lasts 3 seconds out of 600 total seconds (~16.7% true duration), while every minute's first 10-second interval gets marked, yielding 30 scored intervals out of 30 = 100% recorded prevalence. Partial interval always overestimates, especially for brief, infrequent behaviors.

During IOA assessment, Observer A records 8 intervals as '+' and 12 as '−' using whole-interval recording. Observer B records 10 intervals as '+' and 10 as '−'. Their '+' intervals overlap in exactly 6 instances. Which IOA method is most conservative, and what is its value?

Answer: Scored-interval IOA; 6/12 = 50%

Scored-interval IOA (also called occurrence agreement) counts only intervals where at least one observer scored '+', then calculates agreements over that subset: both agreed on '+' in 6 intervals, and the total intervals scored '+' by either observer = 6 + (8−6) + (10−6) = 6+2+4 = 12. So scored-interval IOA = 6/12 = 50%. This is the most conservative because it focuses on the harder-to-agree-on occurrence intervals, excluding easy non-occurrence agreements. Unscored-interval IOA uses '−' agreements only. Exact agreement uses all intervals. Total count IOA compares totals, not interval-by-interval.

A practitioner needs to measure the latency of a client's self-injurious behavior (SIB) in response to demands. However, the client sometimes emits SIB spontaneously before any demand is delivered. Which measurement modification best preserves the validity of the latency data?

Answer: Exclude trials in which SIB occurs before the demand is delivered and restart the trial clock only at demand onset

Latency measures the elapsed time between a specific antecedent (the demand) and the onset of the target behavior. If SIB precedes the demand, no valid latency can be calculated for that trial because the antecedent-behavior temporal sequence is violated. The correct solution is to void that trial and restart timing only when a clean demand is delivered. Recording from session start (B) conflates pre-demand behavior with demand-evoked behavior. Partial interval (C) does not capture latency at all. Frequency (D) is a different dimension and doesn't solve the antecedent-sequencing problem.

A researcher uses a multiple-baseline design and graphs data using a semi-logarithmic (Standard Celeration Chart) scale rather than an equal-interval scale. A behavior that doubles from 2 to 4 responses per minute and another that doubles from 20 to 40 responses per minute will appear visually as:

Answer: Equal-sized slopes for both behaviors, because the chart captures proportional (multiplicative) change

The Standard Celeration Chart uses a logarithmic Y-axis, meaning equal visual distances represent equal ratios (×2, ×10, etc.) rather than equal absolute differences. Both 2→4 and 20→40 represent a ×2 (doubling) change, so they occupy the same vertical distance on the chart and produce slopes of identical steepness. This is a key advantage over equal-interval charts, where 20→40 would look 10× more dramatic than 2→4 even though both represent the same proportional learning rate (celeration).

An analyst is training a new therapist to use momentary time sampling (MTS) at 1-minute intervals to measure on-task behavior. The behavior analyst wants to ensure the MTS data are a valid estimate of true duration. Under which condition does MTS most accurately reflect actual duration percentage?

Answer: When the behavior is randomly distributed and its bout duration is roughly equal to the inter-bout interval

MTS estimates duration by sampling behavior state at a single moment per interval. Its accuracy as a duration estimate is highest when the behavior is randomly (uniformly) distributed across the session and when bout durations approximate inter-bout intervals — conditions under which the sample moments are equally likely to catch occurrence or non-occurrence proportional to true prevalence. Clustered bouts (B) cause systematic over- or under-estimation depending on when sampling occurs. Shorter intervals than bouts (C) reduces missed occurrences but doesn't make MTS estimate duration accurately. Fixed-onset behavior (D) creates a non-random distribution that biases sampling.

A behavior analyst graphs a client's daily rate data and notices the slope across a 10-session baseline is visually flat with low variability. Upon calculating the split-middle line of progress, they find the slope is actually slightly decelerating. What is the most clinically significant implication of this discrepancy?

Answer: Visual inspection of raw graphed data can be misleading; the split-middle line reveals a trend that may indicate the behavior would continue to worsen without intervention even though the raw data appear stable

The split-middle method (a form of celeration line calculation) divides the baseline into two halves, finds the median rate and median time point of each half, and draws a line through those two points. It can reveal a genuine decelerating trend even when individual data points appear visually clustered around a flat line — because subtle downward drift accumulates across sessions. Clinically, a decelerating target behavior baseline means the behavior is trending in the desired direction naturally, making it harder to attribute later improvements solely to intervention (a threat to internal validity). This must be acknowledged in analysis. Option B is incorrect — flat appearance doesn't preclude a mathematically decelerating trend. Option C is partially true but misses the internal validity concern. Option D is a common misconception; variability and trend are independent features.