Estimating actor, partner, and interaction effects for dyadic data using PROC MIXED and HLM: A user-friendly guide LORNE CAMPBELL a AND DEBORAH A. KASHY b a Simon Fraser University; and b Michigan State University Abstract Data collected from both members of a dyad provide abundant opportunities as well as data analytic challenges. The Actor-Partner Interdependence Model (APIM; Kashy & Kenny, 2000) was developed as a conceptual framework for collecting and analyzing dyadic data, primarily by stressing the importance of considering the interdependence that exists between dyad members. The goal of this paper is to detail how the APIM can be implemented in dyadic research, and how its effects can be estimated using hierarchical linear modeling, including PROC MIXED in SAS and HLM (version 5.04; Raudenbush, Bryk, Cheong, & Congdon, 2001). The paper describes the APIM and illustrates how the data set must be structured to use the data analytic methods proposed. It also presents the syntax needed to estimate the model, indicates how several types of interactions can be tested, and describes how the output can be interpreted. People involved in dyadic relationships (or even brief dyadic interactions) often influence each other’s cognitions, emotions, and beha- viors. This notion is certainly applicable to romantic relationships, where the potential for mutual influence may be the quintessential feature of closeness in relationships (Kelley et al., 1983). For instance, virtually all major theories of romantic relationships acknowledge the concept of interdependence, including theories of equity (Messick & Crook, 1983; Walster, Walster, & Berscheid, 1978), commit- ment (Rusbult, 1980), trust (Rempel, Holmes, & Zanna, 1985), interdependence (Kelley & Thibaut, 1978; Thibaut & Kelley, 1959), and attachment (Bowlby, 1969, 1973, 1980). Mutual influence is also germane to other types of dyadic relationships (e.g., friendships and parent-child relationships). One conse- quence of interdependence is that the attributes and behaviors of one dyad member can impact the outcomes of the other dyad member. Relationships researchers have struggled with ways to analyze dyadic data. Although some relationships researchers continue to analyze dyadic data by ignoring the interde- pendence and simply analyze the data as if it were derived from a set of individuals, most relationships researchers are cognizant of the problems inherent in such an approach (i.e., biased significance tests; Kenny, 1995). Con- cern about nonindependence has led many researchers who study dating or married couples to conduct separate analyses for men and women. This circumvents the nonindepen- dence issue, but it produces its own problems. Perhaps the most substantial problem with separate analyses for men and women is the implicit assumption that gender is an important factor, and that differences between men and women exist. All too often when researchers find different prediction equations for men and women, they interpret their results as implying that there are significant differences between 327 Lorne Campbell is now at the University of Western Ontario. Correspondence should be addressed to Lorne Campbell, Social Services Centre, University of Western Ontario, London, ON, N6A 5C2, Canada, or Deborah A. Kashy, Department of Psychology, Michi- gan State University, East Lansing, MI, 48824, e-mail: kashyd@msu.edu. Personal Relationships, 9 (2002), 327–342. Printed in the United States of America. Copyright # 2002 ISSPR. 1350-4126/02