Association arrays for the study of familial height, weight, lipid, and lipoprotein similarity in three West Coast populations.

A more refined understanding of familial similarity may be achieved through a collection of measures of dependence that is sensitive to a variety of nonlinear trends ans stochastic relationships between trait values. Parent-offspring, spouse, and sibling similarities are examined by association arrays that assess dependence between variables for appropriate classes of functions (e.g., the class of all increasing functions). The methodology is applied to height, weight, lipid, and lipoprotein variables collected in nuclear families at the Seattle, Stanford, and La Jolla Lipid Research Clinics. Among the results obtained using association arrays, there is the suggestion that spouse similarity for standardized weight is strongest for functions emphasizing the higher values of the wives' weight independent of the husbands' weight, and that sibling similarity for high density lipoprotein cholesterol concentrations appears strongest for functions emphasizing the higher values of the siblings. The results deduced from the method of association arrays are compared and contrasted with those obtained from standard correlations.