Abstract
Governments serve a variety of purposes, and where governments spend their money has always been of concern to society. In particular, spending on public education is of great interest. However, the volume of this information can be difficult to manage. Therefore, the purpose of this work is to compare the COSTATIS method and generalized Procrustes analysis (GPA) when working with multi-way data. Despite the particular characteristics of each of them, they present similarities and differences that, when analyzed together, can provide complementary results to researchers. The COSTATIS consists of a co-inertia analysis of the compromise of two k-table analyses. The GPA method provides an optimal superimposed representation of individual configurations, and a common consensus configuration is constructed as the mean of all transformed configurations. In addition, the GPA method includes the translation, rotation and scaling of coordinates. In this study, both methods were applied, and the advantages and disadvantages of each are presented. The treated data are a sequence of tables from various countries where different public expenditures on education have been measured over time.
Highlights
Education is a fundamental human right for the achievement of stable relations among people, and it is an essential part of the welfare state
COSTATIS and generalized Procrustes analysis (GPA) methods analyze the relationship between public education expenditure structures of high- and low-nominal GDP countries from 2005 to 2019
Both methods indicate that remuneration of all staff (SCE), expenditure on the tertiary level of education (LTER), and public expenditure on education as a share of total public expenditure (GEE) behave in both high- and low-economy countries
Summary
Education is a fundamental human right for the achievement of stable relations among people, and it is an essential part of the welfare state. Quantifying the efficiency of public expenditure to support education provides the opportunity to accumulate large amounts of data. This is very important because it highlights solutions to improve educational indicators [1]. Datasets can be multidimensional as they are collecting information from objects and variables obtained from various periods (three-way data) (I × J × K). It is interesting to study multiblock data with several hierarchically organized blocks, as there can be partitions due to objects, variables or time. In the treatment of these data, various statistical techniques have been developed that enable a simultaneous analysis of the tables under study by obtaining a consensus structure capable of synthesizing all the available information
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
