Methodological Similarities and Differences in the Assessment of Peer Relationship Dynamics

• In Event: 1-219 - Considerations in the Assessment of Social Context and Peer Network Structure

Integrative Statement

INTRODUCTION: Peers have considerable influence over one another. This influence may occur at the dyadic, group, or social network level. To reliably assess peer influence data, however, social contextual effects must be accounted for. Innovative analyses have emerged to do so. For example, dyadic data analysis (APIM; Kenny, Kashy, Cook, 2006), group level analysis (MLM; Goldstein, 1987, SEM; Bollen, 2005), and longitudinal social network models (RSiena; Ripley, Snijders, Boda, Voros, & Preciado, 2018) are used to assess various levels of social context.
GOALS: We set out to assess the role of social context in adolescent adjustment outcomes using three statistical approaches. In these illustrations, we consider the similarities, differences, strengths, and limitations of each of the various methods for assessing social contextual effects. Additionally, we outline recommendations for when (and under what conditions) each method might be used.
METHOD: Two longitudinal datasets will be used throughout these illustrations. The first dataset is a sample of 1,995 3rd and 4th grade Finnish students (46% girls; ~10 years). The second dataset used in this illustration is a sample of 9,434 2nd, 3rd, and 4th grade Dutch students (50.4% girls; ~9 years). Peer nomination data was coded. From these data, dyadic level data was created of stable friendships, group level data was created of social norms and hierarchy structures, and social network data was created using all peer affiliate nominations.
ANALYSIS: The current investigation demonstrates three innovative methodologies used to assess the impact of social context on child adjustment outcomes. Particularly, the models used have been championed in assessments of peer influence that are free from the statistical bias that results from non-independence issues and other external social confounds (e.g., social norms, peer group structures). The Actor-Partner Interdependence Model (APIM; Kenny, Kashy, & Cook, 2006) partitions variance within and between dyadic interaction partners. Multilevel models partition variance between individuals and groups. Simulation Investigation for Empirical Network Analysis (SIENA; Snijders et al., 2010) longitudinally and simultaneously models selection and influence effects, while also simultaneously accounting for features of social network structure and network norms.
RESULTS: The results of model illustrations and comparisons illustrate both similarities and differences across methodologies. For example, all methods can be used to assess forms of peer influence. The difference in these methods is primarily in the flexibility of the assessment of social contextual effects. Perhaps the most dynamic and flexible estimates of social contextual effects are found within longitudinal social network models; however, the result of these various model estimates and comparisons likewise indicate that there may be some potential advantages of more traditional analytic techniques (e.g. APIM, MLM, SEM). Thus, in general, it is critical to match statistical method to research goals.
CONCLUSION: Similarities and differences as well, as the potential strengths and limitations, of each approach will be discussed. Illustrations will be provided. Discussion center around the importance of matching statistical methodology to research goals, as well as the importance of reliable and unbiased assessment.

Authors

• Dawn DeLay, Arizona State University
  Presenting Author

• Adam A. Rogers, Brigham Young University
  Non-Presenting Author

• Brett Laursen, Florida Atlantic University
  Non-Presenting Author

• Noona Kiuru, University of Jyväskylä
  Non-Presenting Author

• Thomas A Kindermann, Portland State University
  Non-Presenting Author

• Rene Veenstra, University of Groningen
  Non-Presenting Author

• Jari-Erik Nurmi, University of Jyvaskyla
  Non-Presenting Author