Developmental Epidemiological Framework for Family Research on Depression and Aggression
Sheppard G. Kellam, M.D. Johns Hopkins University
Recent advances in family research have caused investigators to realize the need for theoretical and methodological integrations and a wider interdisciplinary perspective. More limited perspectives have produced fairly high levels of precision in measurement and hypothetical models to explain outcomes, but new problems have arisen based on these successes. Current sampling procedures often leave uncertain the populations of individuals or families for whom research findings pertain. In many family studies the sample is not representative of a defined population. Further, the frequency and distribution of family processes that put children and families at increased risk cannot be measured without defining the population under study. Most important, causal models are frequently limited by not defining the population and including in the model relevant aspects of the environmental contexts with which the family and its members must articulate.
Prior family research has frequently been done on volunteer populations or populations drawn from clinics where the families have sought help for problems. Such samples come from unknown total populations and entail selection bias in the sampling, since those families who seek help are likely to be importantly different from families with similar problems who do not seek such help ( Greenley & Mechanic, 1976; Greenley, Mechanic, & Cleary, 1987; Kellam, Branch, Brown, & Russell, 1981). Those who seek help from the church may be quite different from those who seek help from the clinic; many families do not seek help at all. Relying on volunteer subjects has similar problems when they are sought through newspaper or poster advertising; those that respond may not