Foundations
What every APIM analyst needs to know before opening R
Before you fit your first model, you need to know why the APIM exists. The four pages in this section give you the conceptual grounding. They are short on purpose — read them in order, and then move on to the tutorials.
Familiarity with multilevel (random-effects) models and structural equation models. In R, basic working knowledge of the lme4 and lavaan packages.
The four concepts
1. Non-independence
Why two people in the same dyad violate the standard regression assumption of independent residuals — and why this matters for every downstream test.
2. Distinguishability
The single most important design decision in dyadic analysis. Are the two members of each dyad interchangeable, or is one of them marked by gender, role, or time?
3. k-patterns
Four ways partner and actor effects can combine — actor-only, couple, contrast, and mixed. How to test which pattern your data fit.
4. Data structures
Long format (one row per person) vs. wide format (one row per dyad). When to use which. The general “doubling” rule.
A 60-second recap
Two people in a dyad share an environment, a history, and a future. Their scores are correlated for reasons that have nothing to do with causation. Standard regression ignores this and gives you standard errors that are too small. The Actor–Partner Interdependence Model fixes this by pooling the dyad as the unit of analysis (so each dyad contributes one piece of information, not two) and by decomposing every effect into an actor part and a partner part. The model is the same whether your dyad is indistinguishable (siblings, strangers paired in an experiment) or distinguishable (romantic partners, parent and child). What changes is the constraint you impose on the actor and partner slopes.
Once these four ideas are clear, the eight tutorials in this site should feel like variations on a single theme rather than eight unrelated recipes.