The GSMS is a longitudinal study of the development of psychiatric disorder based in a predominantly rural area of the southern United States. As part of an accelerated cohort (Schaie, 1965) two phase sampling design, a representative sample of 4500 9, 11and 13 year-olds resident in western North Carolina was selected using a household equal probability design from a population of approximately 20,000 children. In the screening phase parent report data addressing behavioral disorder items from the Child Behavior Checklist (CBCL, Achenbach & Edelbrock, 1983) were collected. The interview phase included all children scoring above a pre-defined CBCL score (designed to identify the highest scoring 25% of the population) and a 10% random sample of the remainder. The GSMS target population includes an American Indian reservation and all age-appropriate American Indian children were recruited. American Indians formed 25% of the sample although they constituted 3% of the study area population. In order to generalize results back to the target population, observations were weighted proportionately to the inverse of their selection probability. Participation included 1420 individuals (56% male); 508 aged 9,
