Abstract:

The great territory and different social and economic conditions of Russian regions causes the variety of characteristics for more than 40 000 Russian schools. For many years, we can observe the significant difference between the results of the international, national and regional education quality researches (for all levels of the general education system).

The correct comparison of the education assessment results makes a big difference for the design the programs of the "low-efficiency" schools support and on the other hand – identification of the schools, which can show the best results in adverse social conditions.

Authors developed the method of collecting the data about outer factors, which have a significant impact to the school educational results. The list of these factors was formalized, the data sources were defined (statistical observation forms, primary data from information systems, etc.). We suggest the methods to process and cleaning the data, the outer factors aggregation approach. In addition, we solve the problem of comparison of the education assessment results on a single scale. For example, in one of the subjects of Russian Federation (Tomsk region) we collected the data for more than 120 000 students and their families from 304 schools (97% – 119 500 students, after processing and cleaning of the primary data).

In this paper, we justified the school grouping by categories: town, country, ungraded (country) schools. Using the Fisher and Kolmogorov-Smirnov criteria, we proved statistically significant differences between these groups (at the significance level α = 0.05) for results of two different educational assessments. Using the regression model, for three groups of schools were identified the context information, which can characterize the school features. Based on these information, we calculated the index of outer school context (IOSC).

Within each group of schools (town, country, ungraded) the method of clustering by IOSC values was introduced. Authors constructed the linear regression models, which show how the results of Final State Exams for 9th grade (FSE-9) depend on IOSC. The regression model for the town schools group can explain the dispersion of 30% of FSE-9 results (30% of FSE-9 results depends on outer school context, without teachers’ and school’s resources contribution).

The proposed method of clustering allows justifying the necessity of different approaches to the town, country or ungraded schools support. In addition, this method allows to compare the educational results within groups of schools with similar outer conditions, which cannot be influenced by school, or identify schools, which show significantly dissimilar from the cluster mean results (resilient and failing).

Therefore, if regional administrators have to design and realize the programs of the "low-efficiency" schools support, they should solve the problems of schools, taking into account their specific outer context (using clusters).

Keywords:

Clustering, design the programs of the "low-efficiency" schools support, regression model, the index of outer school context.