Dyadic Analysis
Also known as: dyad analysis, dyadic data analysis, social relations model, dyad census
Dyadic analysis treats the dyad — the pair of actors and the relation between them — as the unit of analysis, separating the relational outcome into what each actor brings to all their relationships and what is unique to the specific pair. It spans the descriptive dyad census of network analysis and statistical frameworks such as Holland and Leinhardt's p1 model and Kenny's Social Relations Model, all of which respect the structural non-independence inherent in relational data.
Key highlights
- Respects the structural non-independence of relational data that standard regression violates.
- Decomposes relational outcomes into interpretable actor (sender/receiver) and dyad-specific components.
- Quantifies reciprocity at both the dyadic and generalized levels.
- Provides a graded toolkit from the descriptive dyad census to full statistical models (p1, SRM, ERGM).
Intuition
This section is available to Pro members. Upgrade to Pro
How it works
This section is available to Pro members. Upgrade to Pro
When to use it
Use dyadic analysis whenever the data are relational pairs and the non-independence of observations sharing an actor must be respected — round-robin ratings, directed networks, partner studies, family and couple data. The dyad census and reciprocity statistics describe directed structure; the Social Relations Model decomposes round-robin or block designs into actor and dyad effects; the p1 model and its ERGM successors model binary tie formation. It is not appropriate for genuinely independent observations, and the Social Relations Model requires suitable designs (round-robin or block) with enough partners per actor. The p1 model assumes dyads are independent given the actor effects, which ERGM relaxes when higher-order dependence (transitivity) matters.
Strengths & limitations
- Respects the structural non-independence of relational data that standard regression violates.
- Decomposes relational outcomes into interpretable actor (sender/receiver) and dyad-specific components.
- Quantifies reciprocity at both the dyadic and generalized levels.
- Provides a graded toolkit from the descriptive dyad census to full statistical models (p1, SRM, ERGM).
- The Social Relations Model needs round-robin or block designs with several partners per actor to identify components.
- The p1 model assumes dyad independence given actor effects, ignoring transitivity and other higher-order dependence.
- Variance-component estimates can be unstable in small groups.
- Descriptive dyad-census statistics require a null model to judge whether reciprocity exceeds chance.
Common pitfalls
This section is available to Pro members. Upgrade to Pro
Applications
This section is available to Pro members. Upgrade to Pro
Frequently asked
Why can't I just use ordinary regression on dyadic data?
Because each actor appears in many dyads, observations sharing an actor are correlated, violating the independence assumption of ordinary regression and producing badly understated standard errors. Dyadic methods — the Social Relations Model, p1, ERGM, or multilevel models with crossed actor random effects — explicitly model this dependence, separating individual tendencies from dyad-specific effects and giving valid inference.
What is the difference between dyadic and generalized reciprocity?
Dyadic reciprocity is whether a specific pair's unique relational effects are correlated — if Ana has an unusually positive tie to Beto, does Beto have an unusually positive tie back? Generalized reciprocity is an actor-level property — whether actors who direct more ties to others also tend to receive more ties. The Social Relations Model estimates both as separate correlations.
How does the p1 model relate to the ERGM?
The p1 model is the special case of the exponential random graph model that includes only density, actor sender/receiver effects, and a single reciprocity parameter, and that treats dyads as independent given those effects. The ERGM generalizes it by adding terms for higher-order dependence such as transitivity and shared partners, which p1 cannot capture. Historically p1 was the seed from which the ERGM framework grew.
Sources
- 1.Holland, P. W., & Leinhardt, S. (1981). An exponential family of probability distributions for directed graphs. Journal of the American Statistical Association, 76(373), 33–50.
- 2.Kenny, D. A., Kashy, D. A., & Cook, W. L. (2006). Dyadic Data Analysis. Guilford Press.ISBN 978-1-57230-986-9
You have read it. What now?
Cite this page
ScholarGate. (2026, June 22). Dyadic Analysis. ScholarGate. https://scholargate.app/sociology/dyadic-analysis