Performance criteria of learning math tests

In present days modern computer tools and information technologies implementation in educational process are widespread. The main purpose of computer-based testing shifted from the assessment instrument of learning outcomes to the purpose of teaching. Performance criteria of learning mathematical tests are not yet studied properly. The goal of our work is to find and to study the performance criteria of learning math tests. First, we attempted to formulate the performance criteria of learning math tests. We assume these criteria to be informativity, discriminative criterion, validity, and reliability. Thus, we created 21 learning tests based on the course of mathematical analysis. We tested approximately 48 first year students of the physical department of the National Research Tomsk State University (Russia) to check how fully our tests correspond to the purpose of teaching. Then we performed the study of the statistical data, based on the criteria test results. Thus, we found out our tests informative since we used the scale intervals and since the results of testing were easily processed statistically. Due to the fact that the data obtained are distributed uniformly or normally, we concluded that our tests satisfy the discriminative criterion. Our tests are also valid, because they measure exactly what they are intended to measure. Nevertheless, we describe the tests at hand are not always have a high reliability. It can be improved by increasing the test value, but it leads to the loose of student's interest and attention to the task. In the following we are planning to do the reliability improvement without increase in the test tasks value, and to perform the study of normative data availability and criteria established by the experts.