Untangling comparison bias in inductive item tree analysis based on representative random quasi-orders

Inductive item tree analysis (IITA) comprises three data analysis algorithms for deriving quasi-orders to represent reflexive and transitive precedence relations among binary variables. In previous studies, when comparing the IITA algorithms in simulations, the representativeness of the sampled quasi-orders was not considered or implemented only unsatisfactorily. In the present study, we show that this issue yields non-representative samples of quasi-orders, and thus biased or incorrect conclusions about the performance of the IITA algorithms used to reconstruct underlying relational dependencies. We report the results of a new, truly representative simulation study, which corrects for these problems and that allows the algorithms to be compared in a reliable manner.