From a methodological point of view, the study of family relationships entails paying specific attention to the measuring and analysis of relational features. Family studies are characterized by complexities stemming from the different points of view of the family members and the various dyadic relationships involved; likewise, family research is characterized by the issues related to the non-independence of the data collected. The contributions included in this special issue show innovative research designs and statistical techniques that allow the researcher to manage and analyze the non-independence of family data.
In their contribution Lanz, Scabini, Tagliabue, and Morgano identify the pivotal methodological questions that researchers must consider in planning research projects that preserve the relational specificity of the family, taking into account non-independence among family members and within relationships. Their review of the field of family studies points out that, despite the progress that has been made regarding the methodological ability to handle relational issues within the
area of family research, few studies either consider more than one family relationship or adopt a family or dyadic unit of analysis.
Claxton, DeLuca, and van Dulmen focus on how to manage the issue of nonindependence of dyadic data when assessing psychometric properties. They examine various ways to account for non-independence of data and recommend the best practices to validate dyadic measures using confirmatory factor analysis (CFA). Indeed, CFA models should be particularly attractive to relationship scholars, as they
enable researchers to model non-independence inherent to dyadic data and to test whether measures work similarly across dyad members.
Tagliabue and Donato explore how the non-independence of family data affects missingness. Missingness in family research, can be classified into three different levels: item, respondent, and dyad. The authors provide a step-by-step description of the procedures used to analyze the degree of missingness and the mechanisms at the root of it at the various levels.
Luginbuehl and Schoebi illustrate how family processes and individual functioning can be studied empirically as they evolve over time within a family’s natural environments using interpersonal extensions of ambulatory assessment designs (repeated measures designs). These interpersonal extensions refer to family processes that can be assessed at the within-dyad level, unfolding within dyads across time and situations. A multilevel analytic approach has been used to handle the various
sources of non-independence.
In her contribution, Barni focuses on relative importance analysis, a fairly new set of techniques which can estimate the unique and shared contribution of one predictor within the context of the other predictors. This range of techniques is extremely fruitful for those family studies in which the measures of family-related variables are often moderately to highly correlated with one another. In fact, when data can be gathered from more than one family member or can be used to asses
more than one family relationship, it is common to see high correlations among family variables.
Alfieri and Lanz propose a multidimensional measurement model to be used in the exploration of specific characteristics of family relationships. These authors adopt a multiple perspective and a multi-relationship approach to investigate the family construct in order to develop measurement models that reflect the complexities of the family. In particular, in these measurement models individual, dyadic, and family levels are identified. Confirmatory factor analysis is used to compare different measurement models.
The complexities of the family relationship can be used to analyze the social relations model (SRM) with only one family member, as illustrated by Cook in his contribution. Cook also addresses the importance of the use of the SRM with only the mother for the clinical assessment of families. This innovative application of the SRM is an important tool for clinical assessment, as it allows researchers to identify the sources of variance in the subjective experience of the family system.
Paleari and Fincham present cross-lagged latent difference score (LDS) models as a method for effectively conceptualizing and measuring change in twowave dyadic data. These models allow us to estimate and describe within-person changes as well as their covariation across partners. The cross-lagged LDS models are shown as being complementary to, rather than competitive with, other betterknown methods of analysis.
The articles comprising this special issue are meant to be used as guides for critical methodological issues and statistical techniques with which some researchers might be only modestly familiar. The special issue gives new insight into how to deal with non-independence of family data without oversimplifying, and without losing the relational specificity of the family.Back