A model [the causal–common–contingent (CCC) model] examining this relationship specifically and testing the degree of overlap between initiation and progression liabilities, as well as their genetic and environmental influences, was developed by Kendler and colleagues [21,22], and has been extended recently [23]. The model was designed to examine contingent data, i.e. where availability of information on a second variable (progression) is dependent on the response to an earlier variable (initiation). Thus initiation must, by definition, occur before individuals can progress to more frequent, heavy or problem use. Information on progression will therefore be available only for those who have initiated use. These types of data cannot be analysed with other more conventional models used frequently in twin studies, such as the Cholesky model or causal model [15].
