Power and Sample Size

Abstract

In this chapter we discuss the power of the twin study to detect variance components in behavioral characters. Our discussion is not in any way intended to be an exhaustive description of the power of the twin study under all possible combinations of causal factors and model parameters. Such a description is in large part available for the continuous case (Martin et al.,1978) and there is an extensive comparison of the power of various designs to detect cultural transmission (Heath et al., 1985). As we move out of the framework of the univariate classical twin study to consider multivariate hypotheses and data from additional classes of relatives, a comprehensive treatment rapidly becomes unmanageably large. Fortunately, it seems rather unnecessary because the prospective researcher usually has certain specific aims of a study in mind, and often has a reasonable idea about the values of some of the parameters in the model. This information can be used to prune the prodigious tree of possible scenarios to manageable proportions. All that is required is an understanding of the principles involved, which we aim to provide in Section 9.2. We illustrate these methods with a limited range of examples for continuous and categorical twin data.