Evaluating normal approximation confidence intervals for measures of 2 × 2 Association with applications to twin data

Twin data are of interest to genetic epidemiologists for exploring the underlying genetic basis of disease development. When the outcome is binary, several indices of 2 × 2 association can be used to measure the degree of within twin similarity. All such measures share a common feature, in that they can be expressed as a monotonic increasing function of the within twin correlation. The sampling distributions of their estimates are influenced by the sample size, the correlation and the marginal distribution of the binary response. In this paper we use Monte-Carlo simulations to estimate the empirical coverage probabilities and evaluate the adequacy of the classical normal confidence intervals on the population values of these measures.