Personal Relationships, 24 (2017), 242 – 264. Printed in the United States of America. Copyright © 2017 IARR; DOI: 10.1111/pere.12176 Analysis of group composition in multimember multigroup data THOMAS LEDERMANN, a MYRIAM RUDAZ, a AND ALEXANDER GROB b a Utah State University and b University of Basel Abstract Data from groups often have a multimember multigroup (MMMG) structure. Examples are two-parent families with a female or male child (three members, two groups), two same-gender and opposite-gender peers of different status (two members, four groups), or gay, lesbian, and heterosexual couples (two members, three groups). To analyze such data, a framework called MMMG actor–partner interdependence model (MMMG APIM) is presented considering group composition. Three models are discussed in detail: the three-member two-group APIM, the two-member four-group APIM, and the two-member three-group APIM. Structural equation modeling and cross-sectional and longitudinal data are used to illustrate the approach. To ease the interpretation of APIM fndings, a proposal of a general classifcation scheme is made. Interpersonal relationships are important for most individuals across the lifespan (Baumeis- ter & Leary, 1995). Some of the most impor- tant relationships in one’s life include those with family members, friends, peers, teach- ers, and coworkers (Antony & Swinson, 1998). When we study such relationships, we may very often be interested in the extent to which partners infuence each other. For example, we might be interested in whether younger sib- lings are more infuenced by older siblings than younger siblings infuencing older siblings or whether same-gender siblings infuence each Thomas Ledermann, Department of Psychology, Utah State University; Myriam Rudaz, Department of Psychol- ogy, Utah State University; Alexander Grob, Department of Psychology, University of Basel. The research used data of The Iowa Youth and Families Project of Rand Conger and colleagues and the American Couples study of Philip Blumstein and Pepper Schwartz. The data were made available through the Inter-University Consortium for Political and Social Research and the Har- vard University. We thank Larissa Troesch for comments on an earlier version of this article and David Kenny for discussion on the analysis of two-member three-group data. Correspondence should be addressed to Thomas Led- ermann, Utah State University, Department of Psychology, Logan, UT 84322, e-mail: thomas.ledermann@usu.edu. other more than opposite-gender siblings. To assess the degree to which two partners infu- ence each other, the actor–partner interdepen- dence model (APIM; Kenny, 1996) has been developed. This widely used model provides an assessment of intrapersonal and interper- sonal effects, called actor and partner effects, respectively. In its most basic form, the APIM consists of a predictor X and an outcome Y , both measured in both dyad members, which might be stress and satisfaction in husbands and wives. When we study dyads, there is often not just one type of dyad but multiple types. For example, with sibling pairs of different ages, there are four possible group compositions: both male, both female, older sibling male and younger sibling female, and older sib- ling female and younger sibling male. The same is true with friends of different ages or peers of different status. Also, often, there are more than two partners. Examples of triads are two-parent families with a target child or a tutor with two students. As with dyads, there is often more than one type of triad. Two-parent fam- ilies with a child typically represent two dif- ferent groups, one where the child is a male 242