Academic journal article Journal of Marriage and Family

Treating the Dyad as the Unit of Analysis: A Primer on Three Analytic Approaches

Academic journal article Journal of Marriage and Family

Treating the Dyad as the Unit of Analysis: A Primer on Three Analytic Approaches

Article excerpt

This article reviews three analytic approaches for treating the dyad as the unit of analysis. These approaches are useful for the specific, but quite common, situation in which researchers have information from or about two members of a dyad. Three approaches are described: intraclass correlations as a measure of similarity, repeated measures analysis of variance, and hierarchical linear modeling. All three approaches are used to analyze the same data taken from the first wave of a shortterm longitudinal study of 197 families with adolescent children.

Key Words: dyads, statistical analyses.

Family researchers have been encouraged to adopt a dyadic approach to studying relationships (Miller, Rollins, & Thomas, 1982; Thompson & Walker, 1982). This approach to relationships research requires that the dyad be considered as the unit of analysis throughout all phases of the research process from research design to analysis. The first step that many researchers have taken in heeding this call is obtaining information from more than one person in a relationship. Increasingly, researchers collect information from multiple family members on constructs such as personal qualities, behavior toward one another, and evaluations of relationships (Wampler & Halverson, 1993). Once data are collected, researchers confront the challenge of analyzing such rich data. Unfortunately, many interesting questions about relationships are not answered adequately because inappropriate statistical methods are used. This article reviews three analytic approaches that enable researchers to treat the dyad as the unit of analysis. These approaches are intraclass correlations used as an index of couple similarity, repeated measures analysis of variance, and multilevel modeling. These approaches are appropriate for the specific but common situation in which researchers have information from or about both members of the dyad.

Researchers interested in personal relationships have asked some questions that involve the dyad as the unit of analysis and some that involve the individual as the unit of analysis. As Thompson and Walker (1982) explain, some properties that are associated with relationships are individual in nature, for example, love and commitment. These individual properties often are studied with the individual as the unit of analysis. Researchers may ask, "Are people who feel more love for their partner also more committed to their relationship?" Traditional analytical approaches are appropriate in this type of research. However, these individual properties also may be conceptualized in a dyadic way. Researchers may ask, "Are couples who are more in love less likely to divorce?" Thompson and Walker also describe relational properties or properties of the dyad that reflect interdependence or a pattern between members (e.g., power).

These properties can only be studied dyadically. The unit of analysis must be determined by all phases of the research process, starting with how the research question is framed. A researcher can answer questions about the dyad when he or she has information about individual properties. The analytic approaches reviewed here are suited for this situation: when a researcher has collected data on individual properties but has asked questions that are dyadic.

The same constructs can be examined using both individual-level and dyadic-level analyses. Why, then, use a dyadic approach at all? To draw conclusions about the relationship itself, the dyad must be used as the unit of analysis. If conducting analyses at the individual level, a researcher can draw conclusions only about the ways that individuals behave in relationships. This approach is useful for many questions. It is essential that the research question and the approach to data analysis are at the same unit of analysis. If they are not, the researcher risks the ecological fallacy and can draw erroneous conclusions (Robinson, 1950; Thompson & Walker, 1982). …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.