Skip to main content

Table 1 Specification of the multi-level actor-partner interdependence model

From: Actor-partner effects associated with experiencing intimate partner violence or coercion among male couples enrolled in an HIV prevention trial


Predictor labels


Individual actor and individual partner (i)

X_actor, X_partner

Age, race/ethnicity, education, sexual orientation, serostatus, unprotected anal intercourse with a man other than main partner in past 3 months, agreements about sex outside the relationship

Dyad (j)


Duration of relationship with main partner (average); unprotected anal intercourse in last year with main partner; dyadic differences in age, education, race/ethnicity, sexual orientation, and agreements about sex outside the relationship

1. Individual level model:  g‒ 1(mij) = η ij  = β 0j  + β 1j (X _ actor) ij  + β 2j (X _ partner) ij

  Individual level residual term is omitted because its variance is assumed fixed

  η ij is the log odds of the outcome

  β 0j is the within-dyad intercept in dyad j

  β 1j is the slope of ηij on xij in dyad j

2. Dyad level model:  β 0j  = γ 00  + γ 01 (Z) j  + u oj, β 1j  = γ 10 ,  β 2j  = γ 20

  Dyad level slopes are fixed

  u 0j , the random intercept, is the only random effect

  γ 00 is the average intercept across dyads

3. Final model:  η ij  = γ 00  + γ 01 (Z) j  + γ 10 (X _ actor) ij  + γ 20 (X _ partner) ij  + u oj,

  This model contains one random intercept (no random slopes, no interaction terms)