(PS7-76) Evaluating Three Analytic Approaches to Dyadic Ecological Momentary Assessment Data

Abstract Text: Despite an increase in studies that have collected ecological momentary assessment (EMA) data concurrently from two partners (e.g., romantic partners, parent and child), few of these studies have used analyses that truly account for and explore the dyadic structure of the data. This may be due to the inherent complexity of dyadic EMA data in which responses are nested within individuals nested within dyads. Currently, there are no standard approaches or recommendations for analyzing dyadic EMA data. This poster will present three different approaches to examining the same research question: How do parent and adolescent perceptions of parental support and family relationship strength (FRS) predict adolescent sadness and suicidal ideation (SI)?
Data are drawn from an ongoing EMA study of adolescents at elevated risk for suicide and their parents. The sample includes 943 days of data collected from 65 parent-child dyads. At multiple timepoints each day, adolescents were asked to report their perceptions of support received from their parents and FRS using the PROMIS Pediatric Family Relationships Scale (Bevans et al., 2017). Adolescents were also asked about their in-the-moment feelings across a range of affect states (e.g., sadness) and their current degree of passive and active SI. Parents received daily surveys assessing their perceptions of the support they provided their child and overall FRS using the same PROMIS scale.
The three analytic approaches include Gwet’s AC2, actor-partner interdependence modeling (APIM), and multilevel response surface analysis (RSA). A central aim of this poster is to explore the differences in information provided by each approach and to compare what is gained, lost, or interpreted differently if only one were used.

1) Gwet’s AC2 (Gwet, 2014) is a concordance coefficient (range 0.00-1.00) that characterizes the degree of agreement between two reporters. AC2 will be calculated for each predictor for each dyad to quantify within-dyad concordance. Concordance coefficients will then be entered as predictor variables to examine whether lower child-parent agreement on FRS and support are associated with greater sadness and SI. 2) APIM (Kashy & Kenny, 2000; Kenny, 1996) will be used to parse effects into actor and partner effects to determine the degree of influence one member of the dyad’s scores has on the other. 3) RSA (Edwards & Parry, 1993) characterizes the extent to which the combination of two predictive variables relate to one outcome variable . Multilevel RSA will be used to estimate the effects of concordance and discordance between parent and child reports on the two predictors as well as the direction of discordance. Multilevel RSA extends traditional RSA by characterizing how consistent the identified relationships between our predictors and outcome variables are between dyads in the sample.
By comparing and contrasting three different analytic approaches, this poster will provide valuable information to help researchers determine which analyses may be best suited to answer their specific research questions and to take advantage of the inherent richness and complexity of dyadic EMA data.