In most helping professions, progress is assessed subjectively — therapists notice that clients "seem better," teachers feel that students are "more engaged." ABA is fundamentally different: progress must be measured, graphed, and evaluated with scientific discipline. PSY5063 teaches the complete data cycle — from choosing what to measure and how, through graphing and visual analysis, to making and documenting evidence-based treatment decisions.
Measurement in applied behavior analysis
Before data can be analyzed, it must be collected systematically. The first step is defining target behaviors in operational (measurable, observable) terms — writing operational definitions that are precise enough that any two trained observers would agree on whether the behavior occurred. Inter-observer agreement (IOA), also called inter-rater reliability, is calculated to verify measurement accuracy; ABA standards typically require IOA ≥80% before data are considered reliable.
PSY5063 covers the major measurement dimensions and corresponding recording methods:
- Frequency/count: Number of times a behavior occurs in an observation period. Rate = frequency ÷ time. Best for discrete behaviors with clear beginnings and ends (e.g., number of words spoken, hand-washing steps completed, aggressive incidents per session).
- Duration: Total time a behavior occurs during an observation. Best for behaviors defined by their continuation — time on-task, duration of tantrum, length of social interaction.
- Latency: Time elapsed from a discriminative stimulus (SD) to the onset of the behavior. Used when the goal is response speed — time from instruction to compliance, time from alarm to independent morning routine initiation.
- Inter-response time (IRT): Time between the end of one response and the onset of the next. Used to thin schedules of reinforcement or assess rumination frequency.
- Trials to criterion: Number of discrete trials required to achieve a mastery criterion — useful in skill acquisition programs using discrete trial training (DTT).
- Whole-interval recording: Behavior is scored only if it occurs throughout the entire interval; underestimates behavior occurrence. Best for behaviors you want to increase (on-task behavior).
- Partial-interval recording: Behavior is scored if it occurs at any point during the interval; overestimates behavior occurrence. Best for behaviors you want to decrease (self-injurious behavior).
- Momentary time sampling: Behavior is scored only if it occurs at the moment the interval ends. Best for high-frequency behaviors difficult to measure continuously.
Graphing single-case data
ABA uses the equal-interval line graph (not bar graphs or pie charts) because it displays change over time in a way that allows visual analysis of trends and variability. Key graphing conventions include: session numbers or dates on the x-axis; behavior dimension (rate, duration, percentage) on the y-axis; phase change lines (vertical dashed lines) separating baseline from intervention phases; phase labels above the graph identifying each condition; data points within a phase connected by lines, with no line crossing phase change lines.
PSY5063 covers graph construction for multiple single-case experimental designs, including: AB designs (baseline + intervention); reversal/ABAB designs; multiple baseline designs across participants, settings, or behaviors; alternating treatment designs (ATD); and changing criterion designs.
Visual analysis: the four-factor framework
What behavior analysts examine in each phase of a graph
- Level: The mean or median of data within a phase. What is the typical performance? Has level changed substantially between baseline and intervention phases?
- Trend: The direction and steepness of data within a phase. A trend line (calculated by the split-middle method or least-squares regression) shows whether behavior is increasing, decreasing, or stable over time. Trend direction and magnitude are both clinically meaningful.
- Variability: The spread or scatter of data points around the trend line. High variability within a phase signals instability — perhaps the intervention is being implemented inconsistently, or the behavior is sensitive to unmeasured variables. Stable data are necessary for confident interpretation.
- Overlap: The proportion of data points in the intervention phase that fall within the range of baseline data. Low overlap (few intervention data points within the baseline range) supports a functional relationship between the intervention and behavior change. High overlap suggests the intervention may not be effective. The PND (percent non-overlapping data) statistic quantifies overlap.
Data-based decision rules
PSY5063 trains students to apply systematic decision rules rather than making subjective judgments about treatment progress. Common frameworks include the Touchette et al. (1985) decision rules and standard celeration chart (SCC) rules: if three consecutive data points are trending in the wrong direction, make a treatment modification; if three data points meet or exceed the aim line (predicted trajectory to goal), consider advancing to the next phase or reducing support. Decision rules prevent both premature program changes (reacting to random variability) and delayed changes (continuing ineffective treatments too long).
PSY5063 assignments include graph construction, visual analysis write-ups, and data-based program modification papers
Our ABA specialists deliver precisely constructed graphs and accurate visual analysis narratives aligned to BACB standards.
Get Help With PSY5063
Graph construction, visual analysis reports, data interpretation papers, IOA calculations, decision rule applications.
Place Your OrderView All ServicesRelated courses
Frequently asked questions
The debate over visual analysis versus statistical analysis in single-case research has continued since the 1970s. The primary arguments for visual analysis are: (1) behaviorally and clinically important effects should be large enough to see — if an effect is so small that only a statistical test can detect it, it may not be practically meaningful; (2) visual analysis is sensitive to the distribution-free nature of single-case data, which often violate the assumptions (normality, independence of observations) required for parametric statistical tests; and (3) visual analysis encourages direct examination of the data in context, rather than collapsing the temporal record into a single p-value. Critics of visual analysis note that it is influenced by viewer training, graph scale, and context variables. Modern ABA science uses both approaches — visual analysis for applied decisions and for transparency in publication, and effect size statistics (Tau-U, IRD, PND, PAND) for meta-analytic aggregation across studies. In PSY5063, mastery of visual analysis is foundational because it is the primary tool practitioners use daily in program monitoring.