Confidence Interval Estimation for a DifferenceBetween Two Dependent Intraclass CorrelationCoefficients With Variable Class Sizes

摘要

Four confidence intervals for a difference between two dependent intraclass correlation coefficients(ICCs) are developed, focusing on applications to family studies. The basic idea adopted here is thatconfidence intervals for a difference between two parameters can be obtained from confidence limitsfor each parameter. Among the four confidence procedures considered, the one based on the inversehyperbolic tangent transformation for a single ICC performed best. Francis Galton's data on humanstature are used to illustrate the methodology.