approach and that formal modeling is required to determine how well these estimates describe the observed data. Also, this approach does not allow for actual testing of different hypotheses, for example, to test whether it is possible to set any of these effects to zero. It is difficult to discern whether the parent/caregiver rating patterns reflect true shared environment or are instead an artifact of rater bias whereby raters are less able to discriminate between the two twins' aggressive behavior and thus inflate the similarities between them, regardless of zygosity. In fact, when the two co-twins are rated by different teachers (e.g., they are in different classrooms), twin correlations are lower for both MZ and DZ pairs for the wider construct of antisocial behavior (Baker et al., 2007).
