4 - Planning research for social relations modeling

Abstract

When the response of one person in the dyad affects the response of the other member, it is essential to partition actor, partner, and relationship effects. If this is not done, these effects are confounded and one cannot know which are determining dyadic behavior. For example, Huang, Yeomans, Brooks, Minson, and Gino (2017) were interested in the role of asking questions in conversations in separate dyadic “speed dates.” Although there were multiple male-female dyadic interactions that produced an asymmetric block design, the components of dyadic behavior were not partitioned. That research claimed the “data support a trait-level model of question-asking behavior.” (p. 12). Because the data were available, Kluger and Malloy (2017) conducted a social relations analysis and came to very different conclusions. We found that individual differences in questioning (i.e., the trait model) were not the sole determinants when the relationship and partner effects of the social relations model (SRM) were estimated. Actor and relationship variances were revealed, and partner variance was the weakest determinant but was nontrivial. The conclusion that questioning is a stable trait was demonstrated to be incomplete when the appropriate social relations analysis was conducted. This example shows the potential for biased inference when dyadic data from multiple interaction designs are not partitioned into sources specified by the SRM. The general principle is that when the components of dyadic data are ignored, inferences can be biased, erroneous, or incomplete.