mstR: An R package to generate multistage testing designs

Multistage testing (MST; Yan, von Davier & Lewis, 2014) has become an important framework of tailored testing. Similarly to computerized adaptive testing (CAT) it proposes an optimal routing of the administered items according to the previous test taker’s responses. However, items are not selected and administered as single units but as modules (subsets). The main goal of MST consists in selecting the optimal path of modules across the successive stages of the test. Unfortunately, if various operational testing programs are nowadays considering MST for practical administrations, there is still very limited access to accurate software that can either treat or generate MST scenarios for research purposes.
In this talk we succinctly present a new package from the R software, called mstR (Magis, Yan & von Davier, 2017). Built in the same spirit of the package catR for CAT designs, it permits to generate repeated response patterns under a predefined MST scenario by providing the set of modules and related item parameters, the number of stages and the connections between modules from successive stages. Several rules for optimal module selection and ability estimation (under IRT framework or based on test scores) are also available.
This talk will mostly focus on the (non-technical) description of package mstR and its main features. An example will also be described as an illustration of this framework.