We examined three issues: (1) properties of the intraclass correlation coefficient (r1); (2) normality of itsztransform; and (3) Fisher's formula for the variance ofz. To do this we simulated 1,000 samples ofr1and ofzfor several values ofp, both for equal and for unequal family sizes. Those samples permitted us to evaluate some of Fisher's remarks on the bias, variance, and normally ofr1Furthermore, they showed us that all simulatedztransforms were normally distributed with variances that agreed with the values calculated from a modification of Fisher's formula. Those last two results affirm the applicability of theztransform and of Fisher's modified formula to studies in human genetics involving unequal family sizes.
展开▼