Exploratory analysis of the patterns of missing data in the UIS education database

Article type: Research Article

Authors: Ibáñez Salinas, Miguel

Affiliations: Institute for Statistics – UNESCO. E-mail: [email protected], [email protected]

Abstract: A crucial quality dimension of the UIS education database is its capacity to present complete information on national education systems. This exploratory study examines this longitudinal database from the point of view of the missing values and completeness. This seems to be the first study conducted on this critical matter. Several analytical tools are used in this research (e.g. binary factors analysis, cluster analysis, multinomial random effects logistic regression), a fact that reflects the complex nature of the patterns of education statistics production. A structure of 5 dimensions or factors (linked to specific parts of the UIS data collection questionnaires) that describes the manner in which countries' education data are produced is discussed. This structure proved valuable in the classification of countries in 5 clusters, where each cluster can be linked to the capacity of countries to produce/report data. The behaviour in time of these factors was also analyzed; the results showed that production of detailed statistics in primary/secondary and tertiary education statistics are decreasing in time. Lastly, evidence about the positive link between improvements in governance indicators and the increase in the production of education statistics is presented.

Keywords: Missing values, multivariate analysis, education database

DOI: 10.3233/SJI-130759

Journal: Statistical Journal of the IAOS, vol. 29, no. 1, pp. 3-10, 2013