Graph

A graph turns records into a visible series

The horizontal axis often shows date, session, week, or opportunity. The vertical axis shows the measure and unit, such as requests per hour, median latency in seconds, minutes of activity, or successes out of eligible opportunities.

Every point should trace to a record. A graph adds visibility and cannot repair an unclear definition, incorrect denominator, or missing observation.

Labels make the display interpretable

Include:

• a descriptive title
• named horizontal and vertical axes
• units and calculation method
• actual observation dates or sequence
• legend for people, settings, or measures
• phase or procedure-change markers
• indication of missing, invalid, or partial observations
• notes for major access, schedule, health, or system changes

Avoid a vertical scale that exaggerates small shifts or compresses meaningful differences. Show zero when it helps interpretation and explain any truncated axis.

Keep raw denominators available

A point at 80% could mean 4 of 5, 8 of 10, or 40 of 50. Put raw counts in the source table, tooltip, point label, or companion display. For count data, retain observation time. For duration, retain valid observed minutes.

Missing sessions should appear as gaps or clearly labeled missing states. Connecting a line through unavailable data can suggest continuity that was never observed.

A fictional graph review

Mateo is a fictional 13-year-old who wants adults to respond to his requests for clarification during a robotics club. The graph displays independent requests per 30-minute meeting across six meetings: 2, 3, 2, 5, 4, and 6.

The first three points have a mean of 7 divided by 3, or 2.3 requests per meeting. The next three have a mean of 15 divided by 3, or 5 requests per meeting. Observation time remains 30 minutes for every point.

The graph also shows partner responses: 6 of 7 requests across the first three meetings and 12 of 15 later. A vertical marker notes when a new visual parts list and partner training both began. The graph shows a level difference and cannot isolate which change, practice, time, or other factor produced it.

Read level, variability, and direction together

Level describes the typical height of points within a phase. Variability describes how much the values differ. Trend describes a direction across time. A rising line with extreme variability may support a different decision from a stable rise.

Review immediacy, overlap, consistency, and the number of observations when comparing phases. One high point can create an apparent slope. A phase with two points offers little evidence about stability.

Mark measurement changes

A new operational definition, observer, interval length, opportunity rule, device, setting, or graph aggregation can change values. Mark the effective date and keep versions distinct. Recalculate old data only when source evidence supports the new rule and the original record remains preserved.

Do not silently replace session points with weekly averages. Aggregation can smooth meaningful variability and hide missing sessions.

Pair the graph with the person’s experience

A graph can show response timing and still miss comfort, autonomy, effort, pain, family burden, communication access, or whether the goal matters. Include self-report, accessible communication, family context, and relevant quality-of-life outcomes.

Keep AAC, interpreters, sensory and mobility supports, breaks, prescribed care, and emergency help available while collecting the underlying data.

Make the graph accessible

Use readable type, sufficient contrast, direct labels, and patterns or symbols that do not rely on color alone. Provide a data table or text summary for screen-reader and low-vision access. Explain technical terms in plain language.

Ask the person and family whether the display is understandable and which comparison they want to see. A simplified family view can link to the complete clinical record rather than removing needed context.

Ask questions before making a decision

Check whether points use the same definition, unit, observation length, support conditions, and denominator. Look for missing observations and phase changes. Ask how much data the decision needs and what other explanations remain plausible.

When a line is automatically smoothed or a trend is calculated, show the original points too. Record the method, window, and software version so reviewers can reproduce the display.

The CASP ABA Practice Guidelines Version 3.0 public summary places evaluation within ABA behavioral health treatment for people diagnosed with autism. It provides high-level scope and does not prescribe a graph type or visual-analysis rule.