Skip to ContentSkip to Navigation
About us Practical matters How to find us prof. dr. M.J. (Matthijs) Warrens

prof. dr. M.J. (Matthijs) Warrens

Professor (adjunct) of Educational Sciences

Similarity measures

In various research applications it is frequently required that individuals or objects are pigeonholed, that is, they are classified into categories. To compare two classifications researchers use so-called similarity and association measures. These coefficients are formulas that express in a number the similarity or association between two classifications. A value of 1 usually indicates perfect similarity, whereas 0 reflects no similarity. In my research I study properties of various types of measures using algebra and analysis. Examples are Cohen’s kappa, weighted kappa, Cronbach’s alpha and the adjusted Rand index.


Educational trajectories in secondary education

Secondary school choice options include choice of track, profiles, and school subjects. The first aim of this project is to explain students’ educational trajectories within the context of the factual, offered and recommended choice options at their school. That is, school policy and/or culture may restrict students’ choice options and influence their decisions. A second aim is to explain students’ educational trajectories within the context of the choice options they are aware of, their school performance and alternative choice options they evaluate as acceptable. The project provides a systematic evaluation of pros and cons of each choice option and makes this information available to students and schools, which can lead to broader sets of perceived choice options among students.



Benchmarking clustering methods

Cluster analysis is widely used for finding groups in data. There are many different clustering methods, developed and used in many different research disciplines. When applied to the same data two clustering methods will generally produce different groupings. There is generally no best grouping and different groupings of the same data may be meaningful. Furthermore, applying cluster analysis requires making various decisions, e.g. selecting a clustering method, distance measure and the number of clusters. Guidelines on how to make these decisions are limited. In my research I study properties of clustering methods. Useful insights can be obtained by systematically varying choice options for a particular clustering method, and by comparing different methods.

IFCS Task Force on Benchmarking

Last modified:25 June 2022 1.32 p.m.