The effect of polytomous item ratio on ability estimation in multistage tests

The aim of the study is to examine the effect of polytomous item ratio on ability estimation in different conditions in multi-stage tests (MST) using mixed tests. The study is a simulation based research. In the PISA 2018 application, the ability parameters of the individuals and the item pool were created by using the item parameters estimated from the dichotomous and polytomous items obtained in the field of reading skills. MST conditions; panel patterns ("1-2", "1-3", "1-4", "1-2-2", "1-2-3" and "1-3-3"), test lengths (20, 40 and 60), routing methods (Number Correct (NC) and Maximum Fisher Information (MFB)) and polytomous item ratio (10%, 30% and 50%). Simulation data, MST pattern and analysis were obtained with the help of WinGen, Cplex and "mstR" package in R Studio program. A total of 108 conditions and 100 replications were examined in the study. As a result of the simulations; RMSE, mean absolute bias and correlation values were calculated. As a result of the research, it is seen that when the ratio of polytomous items in the tests increases from 10% to 50%, the mean absolute bias and RMSE values decrease, while the correlation values increase. As the test length increases, RMSE and mean absolute bias values decrease, while correlation values increase. In terms of routing methods, MFB performed better than the NC routing method. In general, three-stage panel designs gave significantly better results than two-stage panel designs. In 1-2 and 1-4 panel designs, it does not matter which routing method is used.