Use of binary logistical regress for estimation of quality of the adaptive test

World and domestic experience testifies to efficiency of application of tests for certification of experts in various fields of knowledge and activity. Tests are used for self-training and self-checking; they allow receiving estimation of level of knowledge and abilities, objective enough to reveal gaps in training. In high schools and other educational institutions tests are used for introductory assessment, current control of knowledge, reception of offsets and examinations at any forms of training. With the test monitoring system of knowledge, the reliability of a given mark increases considerably. According to experts, the test system of knowledge monitoring increases the probability of objective assessment of students' knowledge by 80-95%. In the existing system of oral final examination it makes 15-18%. In educational institutions computer testing, which provides individual control of knowledge, regularity of its performing full and objective check of knowledge of the learnt material, unity of requirements, conformity to standards, decrease in operating time for the teacher to control the students' knowledge, is widely used. Experts say obligatory quality check of the test, including its reliability on the statistics basis, is not carried out, the definition of its difficulty and an independent expert review is not done. At the best, the structure and the content of questions is reviewed and discussed without an empirical check. The object of research is adaptive tests. The purpose of research is the processes of adaptive testing connected with their quality estimation. Research objective is to work out models of binary logistical regress and a technique of its construction for correct estimation of quality of adaptive tests applied in educational process. To achieve it the following problems must be solved: to consider indicators of quality of the adaptive test; to study features of the method of binary logistical regress and its realisation in package STATISTICA 6.0; to apply the method of binary logistical regress to estimation of quality of the adaptive test and to analyse the received results. The theoretical base of research is the works by domestic scientists devoted to research of the method of binary logistical regress, approved in scoring and medicine. The scientific novelty consists in the following: the method of binary logistical regress for estimation of quality of the adaptive test, calculating probabilities of occurrence of the right answer at all levels of knowledge is adapted, allowing to estimate the tested for a minimum quantity of test tasks more precisely; such criteria of quality of logistical models as reliability, a xi-square and disagreement relations prove the importance of the constructed models; it is offered to interpret the results of classification of supervision for establishing the quantity of guessed answers and the quantity of random errors.

Keywords

Authors

References