Precision Graphing Slides Precision Graphing Standards
The Architecture of Clinical Data
Lesson 1: Visual Analysis & Interpretation
Spot the Errors
100500
Figure 1. Progress
BAD GRAPH
This graph is used to justify a 50% increase in funding. What is wrong with it?
Improper axis scaling
Disconnected data paths
Missing condition labels
Poor figure captioning
Anatomy of an ABA Graph
The Axes
X-axis (Abscissa) for Time/Sessions.
The Data Path
Clear symbols (circles, squares).
Condition Change
Solid vertical lines for major changes.
Technical Communication
The Figure Caption Rule
"The caption should allow the reader to understand the graph without referring to the main text. It is placed BELOW the graph."
Caption Checklist:
Participant ID
Target Behavior
Measurement System
Key Experimental Events
Building in Excel / Sheets
01
Clean the Canvas
Remove gridlines, legend (if single data path), and default titles.
02
The "Invisible" Phase
Insert blank rows between baseline and intervention phases to create data path breaks.
03
Standardize Symbols
Use 8pt marker size and 1.5pt line width for publication clarity.
Hands-On Workshop Next
Bring your datasets!
Graph Repair Worksheet Graph Repair Workshop
Visual Analysis & Interpretation of Behavioral Data
Student Lab
NAME: ___________________________
DATE: ___________________________
The Challenge
Below is a "bad" graph that violates several ABA and APA publication standards. Your task is to identify the errors and then sketch the corrected version of the graph in the space provided. Precision is critical for accurate visual inspection.
Exhibit A: The Faulty Graph
1000
Fig 1. Behavioral Progress Report
"The data shows a rapid decrease in screaming after the weighted vest was introduced."
1. Standards Violation List
Identify at least five (5) distinct violations of precision graphing standards found in Exhibit A.
1
2
3
4
5
2. The Repair Sketch
Redraw the graph following all standards. Use appropriate scale (0-10 on Y, 1-10 on X). Ensure you represent the two phases: Baseline (Sessions 1-4: data 8, 7, 7, 8) and Weighted Vest (Sessions 5-10: data 3, 2, 1, 2, 1, 0).
Sessions
Frequency of Screaming
Figure Caption:
Rationale for Scale:
Self-Check: Did you disconnect the data path at the phase line? Is the phase line solid? Are condition labels centered?
Dimensions of Visual Inspection Slides Dimensions of Visual Inspection
Going Beyond the Mean
Lesson 2: Interpreting Behavioral Data
Why "Average" Fails
Mean = 10
Stable Level
Mean = 10
Improving Trend
Mean = 10
High Variability
All three have the same mean. Would you treat them the same?
01. Level
The Value on the vertical axis around which a series of behavioral measures converge.
Key Questions:
What is the mean or median level?
Is there a significant shift between phases?
Is the level "high," "moderate," or "low"?
Baseline Level
Intervention Level
Looking for a clear vertical gap between phase data.
02. Trend
The overall Direction taken by a data path.
Direction
Increasing, Decreasing, or Zero (Flat).
Magnitude
Steep vs. Gradual change over time.
Contra-therapeutic Trend
Therapeutic Trend
Note: "Therapeutic" depends on whether the goal is to increase or decrease the behavior.
03. Variability
The frequency and degree to which multiple measures of behavior Yield Different Outcomes.
The Stability Rule:
High variability often requires more data points before a phase change can be justified. It suggests poor Experimental Control.
"Bounce"
Predictability vs. Chaos
Technical Reporting Template
"Baseline data showed a high level with an increasing trend and low variability. Upon introduction of the FCT intervention, there was an immediate level shift to a low level, with a stable zero-trend and low variability maintained across 5 sessions."
This is the standard for graduate work.
Dimensions Identification Lab Dimensions Identification Lab
Visual Inspection & Technical Reporting
Student Practice
NAME: ___________________________
Instructions
Analyze each graph below. For each phase, describe the Level (Low, Moderate, High), Trend (Increasing, Decreasing, Zero), and Variability (Low, Moderate, High). Then, describe any significant changes between phases (Level Shift or Trend Change).
Case 1: Self-Injurious Behavior (SIB)
Baseline
Helmet
Baseline Analysis
Level: __________________
Trend: __________________
Variability: ______________
Intervention Analysis
Level: __________________
Trend: __________________
Variability: ______________
Description of Change
Case 2: On-Task Behavior
Baseline
Token Economy
Baseline Analysis
Level: __________________
Trend: __________________
Variability: ______________
Intervention Analysis
Level: __________________
Trend: __________________
Variability: ______________
Description of Change
Case 3: Vocabulary Acquisition
Baseline
DTT
Baseline Analysis
Level: __________________
Trend: __________________
Variability: ______________
Intervention Analysis
Level: __________________
Trend: __________________
Variability: ______________
Description of Change
Key Reminder: Always consider the clinical significance of a change. A "statistically significant" level shift may not be "clinically significant" if the baseline level was dangerously high.
Calculated Progress Slides Calculated Progress Analysis
Statistical Aids for Visual Inspection
Lesson 3: Objective Data Evaluation
The "Eye Test" Limit
Visual analysis is powerful but subjective.
Visual Analysis Risks:
Confirmation Bias
Influence of outliers
Low inter-rater reliability
"Statistical aids are used to supplement, not replace, visual inspection in single-subject research."
PND
Effect Size
Split-Middle
Trend Logic
PND
Effect Size
The Calculation Protocol:
Find the Highest (or lowest for reduction) baseline point.
Count how many intervention points exceed that baseline extreme.
Divide by total number of intervention points and multiply by 100.
Scoring PND:
90% + Highly Effective
70% - 90% Effective
50% - 70% Questionable
Below 50% Ineffective
Split-Middle Line
A mathematical way to represent the Trend while accounting for variability.
Process:
1. Divide data into two equal halves.
2. Find the intersection of median session and median value for each half.
3. Draw a line connecting those two intersections.
4. Adjust (up or down) so equal number of points fall above/below the line.
Objective Trend Modeling
Applying the Tools
The Problem:
A student has a high baseline of disruption. The intervention shows a slight downward trend, but the data points are overlapping with the baseline range.
PND Result: 55%
Visual Analysis says "Maybe," but PND says "Questionable." What do you do clinically?
Clinical Debate
Calculated Progress Worksheet Statistical Aids Lab
Calculating PND & Split-Middle Trend Lines
Technical Mastery
NAME: ___________________________
1
Percentage of Non-Overlapping Data (PND)
Case Study: Reducing Aggression
A behavior analyst is evaluating the effectiveness of a Functional Communication Training (FCT) procedure. The goal is to Decrease aggression.
Baseline Data (5 sessions)
12, 15, 11, 14, 13
Intervention Data (8 sessions)
10, 8, 12, 5, 4, 11, 3, 2
Calculation Steps:
A. Baseline Extreme (Lowest baseline point):
B. Count of Intervention points BELOW baseline extreme:
C. Final PND Percentage (B / 8 * 100):
%
Effectiveness Interpretation
Based on the Scruggs & Mastropieri (1998) standards, how would you classify this intervention?
Highly Effective (> 90%)
Effective (70% - 90%)
Questionable (50% - 70%)
Ineffective (< 50%)
2
Split-Middle Line of Progress
Dataset: Daily Social Initiations
Goal: Increase initiations. Data points (Sessions 1-10):
2, 3, 2, 5, 4, 6, 8, 7, 9, 10
Step 1: Divide
Identify First Half (1-5) and Second Half (6-10).
Step 2: Medians
Find the median value for each half.
M1: _____
M2: _____
Step 3: Plot & Connect
Mark the intersection points (Session 3 & Session 8 medians) and draw the line.
Step 4: Adjust
Move the line up or down so half of all points are above and half are below.
Critical Analysis
Does the split-middle line confirm the visual impression of an accelerating trend? Why is this line better than just a "best fit" visual line?
1050
110
Frequency
Sessions
Use a ruler to construct your split-middle line.
Progress Monitoring Protocol v.1.0
Remember: High PND does not guarantee social validity.
Experimental Design Slides Experimental Design Logic
Identifying Functional Relationships
Lesson 4: Scientific Validity in Practice
Did YOU Cause the Change?
In behavioral support, "it worked" is not enough. We must demonstrate Experimental Control.
Functional Relationship:
"A specific change in the dependent variable (behavior) can be produced by manipulating the independent variable (intervention)."
Threats to Validity:
Maturation
History (External events)
Testing Effects
Instrumentation Drift
The AB Design
Baseline + Intervention
A simple comparison between two phases.
Verdict:
Cannot rule out external variables. It is a quasi-experimental design at best.
"I started the vest, and he stopped screaming. But maybe he just got used to the classroom?"
Phase A → Phase B
Reversal Design
A-B-A-B
The Gold Standard for demonstrating control.
Logic of Reversal:
1. Prediction (Baseline)
2. Verification (Re-introduction of baseline)
3. Replication (Re-introduction of intervention)
Warning:
Do NOT reverse if behavior is dangerous or if intervention causes irreversible learning.
A → B → A → B
Multiple Baseline
Staggered Start
Demonstrates control across People, Settings, or Behaviors.
"If behavior only changes when the intervention is introduced at different times for different subjects, we have evidence of control."
Best For:
Irreversible skills (e.g., reading).
Subject 1
Subject 2
Subject 3
The "Stepping Stone" Effect
Choosing Your Weapon
Design Best For... Weakness AB Preliminary data / Clinical pilot No experimental control ABAB Clear functional relationships Ethics of withdrawal MBL Teaching new skills / Safety Baseline must be stable for long
Experimental Design Matrix Design Selection Matrix
Evaluating Experimental Control & Internal Validity
Technical Guide
NAME: ___________________________
Design Feature Withdrawal (ABAB) Multiple Baseline Alternating Treatments Baseline Requirement
|
|
| Proof of Control |
|
|
|
| Primary Limitation |
|
|
|
Design Selection Scenario
"You are working with a student, Eli, who is learning to decode multi-syllabic words. Once Eli learns a word, he does not 'unlearn' it. You want to prove that your phonics-based intervention is responsible for his sudden jump in reading scores."
1. Which design should you NOT use?
Why is it logically impossible in this case?
2. Which design is most appropriate?
How will you stagger the baseline to show control?
Pro-Tip: In Multiple Baseline designs, ensure that the behaviors or participants are functionally independent. If intervention on participant 1 causes change in participant 2 before their intervention begins, you have lost experimental control.
Clinical Decision Slides Clinical Decision Protocol
Data-Based Decisions in Real Time
Lesson 5: The Analytical Capstone
The Practitioner's Duty
Visual analysis is not just for research papers. It is the Compass for your daily clinical practice.
The Decision Dilemma:
If you continue an ineffective intervention, you waste the client's time. If you stop an effective one too early, you lose progress.
STOP & MOD
STAY THE COURSE
FADE & GENERALIZE
STAY
When to continue as-is:
Therapeutic trend is established.
Variability is decreasing/stable.
Goal has not yet been met.
Steady Progress
MODIFY
When to change protocol:
Zero trend over 5+ sessions.
Contra-therapeutic trend.
Excessive, unpredictable bounce.
The "Flatline"
TRANSITION
When to fade support:
Performance consistently meets criterion.
High stability at target level.
Generalization probes are positive.
Mastery Achieved
Rapid Fire Analysis
STOP
GO
You will be shown 10 graphs for 15 seconds each. You must commit to a clinical decision and justify it using Level, Trend, and Variability.
Clinical Decision Simulation Worksheet Clinical Decision Matrix
Applied Visual Analysis & Progress Monitoring
Capstone Simulation
NAME: ___________________________
Scenario
You are the Lead Behavior Analyst. For each client scenario below, review the data and select one clinical decision: STAY (Continue protocol), MODIFY (Change protocol), or FADE (Begin transitioning). You must provide a technical justification based on Level, Trend, and Variability.
CASE A
Marcus: Peer Interactions
GOAL
Sessions 1-5 (Intervention)
Clinical Decision:
Stay
Modify
Fade
Technical Justification
CASE B
Sarah: Out of Seat
Sessions 1-9 (Intervention)
Clinical Decision:
Stay
Modify
Fade
Technical Justification
CASE C
Leo: Manding for Help
CRITERION
Sessions 15-20 (Intervention)
Clinical Decision:
Stay
Modify
Fade
Technical Justification
Peer Review Requirement
Trade matrices with a partner. Do they agree with your justifications?
Behavioral Research Lab Standard v2026.1
