A PND PEM NAP comparison applies each metric to the same valid phases and improvement direction. PEM uses the baseline median, PND uses the improvement-direction baseline extreme, and NAP scores all baseline-intervention pairs. Their values can differ because their reference rules differ. Report all formulas, ties, raw values, trend, and visual-analysis context.
Keep the dataset fixed
Use the same valid phase observations, improvement direction, and missing-state decisions for every metric.
Name each reference rule
PEM compares with a median, PND with one extreme, and NAP with every cross-phase pair.
Explain divergent values
Different results can be mathematically correct. Identify which points and ties drive the divergence.
Avoid metric shopping
Choose metrics for the question and analysis plan, not because one produces the preferred percentage. Preserve visual analysis and repeated demonstrations.
Calculate each metric from one locked worksheet
Start with the same valid baseline and intervention values, phase labels, unit, improvement direction, and missing-state rules. Calculate the baseline median and improvement-direction extreme once. Generate the complete pair table for NAP. Record ties explicitly before applying any metric-specific rule.
Place each formula beside its count. PEM reports intervention values beyond the baseline median. PND reports intervention values beyond the most extreme baseline value in the improvement direction. NAP compares every baseline value with every intervention value and applies the declared score to ties.
Check the arithmetic before comparing results
For baseline 2, 4, 6 and intervention 3, 5, 6, 8, PEM is 3 of 4, or 75%, because 5, 6, and 8 exceed median 4. PND is 1 of 4, or 25%, because only 8 strictly exceeds 6.
The 12 NAP pairs contain eight intervention-favorable comparisons, one tie, and three baseline-favorable comparisons. Baseline 2 contributes four favorable pairs, baseline 4 contributes three favorable and one unfavorable pair, and baseline 6 contributes one favorable, one tied, and two unfavorable pairs. With half credit for the tie, NAP is 8.5 of 12, or 70.8% after rounding.
Explain divergence in plain language
A PND PEM NAP comparison becomes useful when the report explains what each reference rule notices. PND can be strongly influenced by one baseline extreme. PEM uses the baseline median and ignores how far intervention values move beyond it. NAP uses every cross-phase pair and can reflect the overall ordering of the two phases. These distinctions explain why the percentages differ without treating one result as automatically correct for every question.
Report phase length, trend, variability, overlap, immediacy, consistency across replications, and client-valued importance beside the metrics. A favorable supplemental result cannot replace design logic or a qualified visual and clinical review.
Frame one phase comparison
The nonoverlap comparison sheet for Omar identifies the phase labels, improvement direction, raw observations, missing-state rule, reference value or line, formula, tie rule, calculation owner, and review date before a supplemental result is interpreted.
Worked example: one dataset, three reference rules
Omar uses baseline 2, 4, and 6 and intervention 3, 5, 6, and 8, with higher preferred. PEM is 3 of 4, or 75%, because 5, 6, and 8 exceed median 4. PND is 1 of 4, or 25%, because only 8 exceeds maximum 6. NAP is 8.5 of 12, or 70.8%.
Audit every reference and pair count
Omar's comparison uses identical phase values and direction for all three calculations. It lists the median, extreme, pair counts, and tie rules beside each result.
Protect client relevance
During Omar's PND, PEM, and NAP on one dataset review, keep AAC, interpreters, mobility, food, water, bathroom use, prescribed care, health support, rest, relationships, and emergency help available. Use accessible communication to ask whether the measured outcome and magnitude matter to the person. A favorable statistic cannot repair an unwanted goal or inaccessible plan.
Use single-case evidence within scope
For Omar's calculation, the BCBA Test Content Outline supplies examination scope for graphing, interpretation, experimental design, and data-based evaluation. The BACB ethics hub and CASP summary provide professional context. The WWC Version 5.0 handbook is an education-research evidence standard, not a clinical authorization rule. A single-case methods review documents PND, NAP, and extended-line formulas. Research on structured visual-analysis aids and decision accuracy across overlap measures supports using calculations as transparent supplements while retaining metric and design limits. ASHA supports continuous AAC access.
Close the analysis review
Review the nonoverlap comparison sheet with Omar and the responsible clinician. Preserve raw data, graph, phase definitions, calculation version, coverage, ties, limitations, client input, decision, owner, and next review date. Reopen the analysis when data, context, access, health, goal, phase, or method changes.
Clinician review checklist
- Do all three metrics use identical phase values and improvement direction?
- Are the median, extreme, pair table, and tie rule visible?
- Can another reviewer reproduce every numerator and denominator?
- Does the printed NAP result reconcile to favorable pairs plus tie credit?
- Does the report explain why the metric values diverge?
- Are visual analysis, design quality, context, and client priorities retained?
- Does the pair ledger reconcile to eight favorable, one tied, and three unfavorable comparisons?
Related resources
- How to Calculate an Extended Celeration-Line Result
- How to Handle Ties in Single-Case Nonoverlap Metrics
- How to Summarize Immediacy Across Repeated Phase Changes
- How to Calculate Nonoverlap of All Pairs
Sources
- Behavior Analyst Certification Board, Ethics Information and Ethics Codes
- Council of Autism Service Providers, ABA Practice Guidelines Version 3.0 public summary
- Behavior Analyst Certification Board, BCBA Test Content Outline, 6th edition
- What Works Clearinghouse Procedures and Standards Handbook, Version 5.0
- Single-Case Design, Analysis, and Quality Assessment for Intervention Research
- Machine Learning to Analyze Single-Case Data: A Proof of Concept
- Statistical Decision-Making Accuracies for Some Overlap- and Distance-Based Measures for Single-Case Experimental Designs
- American Speech-Language-Hearing Association, Augmentative and Alternative Communication