Characteristics of Families-Implications for Statistical Analysis in Family Nursing Research

Abstract

Research dealing with family phenomena of concern to nursing carries inherent methodological challenges, due in large part to special characteristics of families as social units (Copeland & White, 1991; Daly, 1992; Larzelere & Klein, 1987). Shared history and close contact between family members result in measures obtained from more than one member of the same family being statistically dependent. However, families are composed of distinct individuals and measures from different family members cannot be collapsed or "averaged" without losing important data about each person's unique experience. Perceptions and beliefs of family members are strongly influenced by individual values and by roles in families (determined by age, gender, generation, and family history). Also, families adapt and evolve over time. Describing the shape and determinants of change in families is a priority, but measuring change remains very difficult. Since families are embedded in larger social systems, complex methods of data collection and analysis are often required to capture the reciprocal influences of the environment, the family, and the individual on each other. While these aspects of families have implications for all phases of the research process, many of these characteristics pose major challenges for statistical analysis (Schumm, Barnes, Bollman, Jurich, & Milliken, 1985). Some family researchers have turned to special analysis techniques, for example path analysis and structural equation modeling (LISREL), to assist them in making causal links and verifying theories (Biddle & Marlin, 1987). Ultimately, clinicians and policymakers look to nursing research for answers about which factors are associated with (and perhaps even causally related to) desired or undesired outcomes for families. Testing propositions of emerging theories of family adaptation is a major goal in family nursing research (McCubbin, 1991). The "softness" of the variables many family researchers work with, and the impossibility of manipulating most family conditions experimentally, interfere with conclusively linking variables, yet methods like structural equation