To understand age and sex effects on poly-substance use involvement, substance endorsement data were split by age and sex for multigroup IRT model estimation. As the technical limitations of the model being used did not allow for explicitly accounting for the clustered nature of twin data, we chose to treat first- and second-born twins as an additional group designation and test for “birth order effects” by equating parameters across twins. This resulted in IRT parameters (substance locations and discriminations) being estimated for 12 groups total (2 sexes × 3 ages × 2 twins).
