Investigation of the efficiency of multi-stage testing in open and distance learning: A simulation study

The main objective of measurement and assessment is to estimate the latent ability with a small amount of error and great precision. Although many methods have been applied for this purpose from the past to the present, the most widely used measurement tool is testing. These assessments are administered, especially to a high number of students, using multiple-choice question formats. Within the scope of the research, it is aimed to discover the optimum algorithm for the Open and Distance Learning (ODL) system by examining the possible results obtained in the case of applying the test generation method of the Multistage to the open education exams in a simulative environment in different conditions. In the research process, the simulation-based data were analyzed and compared by using Classical Test Theory, Item Response Theory, and Multistage testing conditions (MST-R and MST-S) estimations in terms of standard error values, correlation coefficients, AIC values, and score rank differences. Classical Test Theory was found to be the approach that fit the data the least well, and the IRT 3PL method was found to be the best one with the lowest error rate. The Multistage test presentation-based estimations, differs significantly from the Classical Test Theory-based estimations, giving results very close to the IRT 3PL model. The conclusions drawn from the study's findings demonstrate that the Multistage testing method can improve the measurement precision of any system, including open and remote learning systems, that use multiple-choice test techniques on big number of students.