B0408
Title: The actor-partner interdependence model for longitudinal dyadic data in the SEM-framework
Authors: Tom Loeys - Ghent University (Belgium) [presenting]
Abstract: Many of the phenomena studied by behavioral scientists are interpersonal by definition. While historically behavioral data were mostly gathered on individuals, these are unprecedented times in terms of the availability of high-quality dyadic. When two people interact in a relationship, the outcome of each person can be affected by both his or her own inputs and his or her partners inputs. The Actor-Partner Interdependence Model (APIM) offers an appealing approach to model such data. When one collects repeated measures on dyads, one must not only contend with the non-independence of the members within a dyad, but also the correlation of the longitudinal measures within a dyad member. This can be achieved by either modeling an autoregressive residual covariance structure or by including a lagged dependent variable in the mean structure. The implementation of the first approach is readily available in multilevel software, but is lacking in the SEM-framework. A complication for the second approach lies in noting that the combination of a lagged outcome in the mean structure and a random intercept can lead to biased inference within standard multilevel software due to the exogeneity assumption. Solutions to those statistical and computational challenges are presented.
