2 Continuous Data Analysis

Biometrical analyses of twin data often make use of summary statistics that reflect differences, or variability, between and within members of twin pairs. Some early studies used mean squares and products, derived from an analysis of variance [Eaves et al., 1977,Martin and Eaves, 1977,Fulker et al., 1983,Boomsma and Molenaar, 1987,Molenaar and Boomsma, 1987], but work over the past 15 years has embraced variance-covariance matrices as the summary statistics of choice. This approach, often called covariance structure analysis, provides greater flexibility in the treatment of some of the processes underlying individual differences, such as genotype $\times$ sex or genotype $\times$ environment interaction. In addition, variances and covariances are a more practical data summary for data that include the relatives of twins, such as parents or spouses [Heath et al., 1985]. Because of the greater generality afforded by variances and covariances, we focus on these quantities rather than mean squares.