Same Symptoms, Different Networks
Visual example
flowchart LR P[Similar symptom profile] --> A[Person A: sleep-driven network] P --> B[Person B: worry-driven network]
See Figures for the expanded example.
A central demonstration of person specificity is to identify participants with very similar questionnaire profiles and examine whether their perceived causal structures nevertheless differ.
Rationale
Two people can endorse nearly the same symptoms at nearly the same severity while having very different beliefs about how those problems interact.
This analysis asks:
How much person-specific relational information remains after similarity in conventional symptom profiles has been controlled?
Possible matching approaches
Participants can be matched on:
- total score,
- subscale scores,
- selected-item severity,
- full item profiles,
- or multivariate distance measures.
A stronger test uses full item-profile similarity rather than total-score similarity alone.
Candidate analysis
- Compute similarity or distance between questionnaire profiles.
- Identify highly similar participant pairs or neighborhoods.
- Quantify network dissimilarity within those matched sets.
- Compare observed network heterogeneity with:
- random participant pairs,
- or pairs matched less closely on questionnaire profiles.
Network dissimilarity measures
Possible outcomes:
- edge-presence disagreement,
- weighted adjacency correlation,
- structural Hamming distance,
- rank difference in out-strength,
- difference in dominant driver,
- difference in global metrics.
Extension
The most compelling version links network differences to later outcomes:
Among participants with similar baseline questionnaire profiles, do differences in perceived causal structure predict different trajectories?
This directly connects person specificity to incremental validity.