The Ministry of Education and Science in Moscow, through its Federal Agency of Education and Department of Statistics periodically collects data from higher education institutions about infrastructure across a broad range of indicators such as a) faculty educational background and course-related publications, b) student and graduate enrollment, c) financial information, as well as d) physical aspects of buildings and equipment.  Because these indicators are collected as frequencies and continuous integers (number of students, total money collected from specific sources, number of successfully defended student theses, and so on), amount of collected data is typically enormous, and Department of Statistics has developed aggregation methods to summarize indicators with statistics and indices.  The aggregation procedures, however, are very complicated and meaningfulness of ratios and weighted averages based on them are not always apparent.  Not surprisingly, interpreting the implications of institutional comparisons for policy and planning are oftentimes difficult.  Analytical burden is also increased because separate indicators are not consolidated into factor structures or dimensional constructs.  Although data are aggregated, they are not linearized hence measurement properties of compiled statistics and indices are unknown.

The purpose of this research is to improve higher education infrastructure reports based on frequency summaries by presenting a latent trait approach to data collected by the Department of Statistics.  Although latent trait theory has been successfully applied to frequency data (Bezruczko, 2003), large volumes of continuous integer data (Álvarez, 2005), as well as medical instrument and laboratory values (scaled integers) (Perkins, Hoffman, and Bezruczko, 2008), empirical examples are still rare.  In this report, infrastructure indicators are modeled as questionnaire items, and frequency and continuous integer values are recoded categorically for analysis with a Rasch model for rating scales.    

          Research question.  This research explores the possibility that indicators of higher education infrastructure may coherently define a latent trait with linear measurement properties.  In this context, the following questions were addressed:

1.    Can statistical frequency and continuous integer data be reformulated for categorical analysis with a Rasch model for rating scales?

2.    Do higher education infrastructure indicators after coding and transformation with a Rasch model have linear measurement properties?

Does formulation of higher education infrastructure constructs offer any benefits to policy analysis?

Andrich, D., Sheridan, B., and Luo, G. (1997-2005). RUMM2020: Rasch Unidimensional Measurement Models software and manual. Perth, Australia: RUMM Laboratory. Bezruczko, N. (2003) Breakthrough Measuring Socioeconomic Status. Journal of Applied Measurement, 4, 137-152. Wright, B. D. and Masters, G. N. (1982). Rating Scale Analysis. Chicago: MESA Press. Rasch G. Probabilistic models for some intelligence and attained tests (Expanded edition, with foreword and afterword by Benjamin D. Wright). - Chicago: University of Chicago Press, 1980. - 199 p. Maslak A.A., Karabatsos G., Anisimova T.S., Osipov S.A. Measuring and Comparing Higher Education Quality between Counties Worldwide // Journal of Applied Measurement. -2005. -v.6. -N. 4. -P.432-442. Maslak A.A. Measurement of Latent Variables in Social and Economic Systems: Theory and Practise. - Slavyansk-on-Kuban: Slavyansk-on-Kuban Press, 2007. - 434 p.