Semiparametric modeling of educational performance characteristics

Authors: Marek Brabec
Publication year: 2024
Publication type: Abstrakt
Periodical: International Conference APPLIED STATISTICS 2024 Program and Abstracts
Projects: Social and motivational factors of study and their influence on study results in tertiary education with a focus on technical fields
Bibliographic citation: BRABEC, Marek. Semiparametric modeling of educational performance characteristics. In: KASTRIN, A., LUSA, L., eds. International Conference APPLIED STATISTICS 2024 Program and Abstracts. Ljubljana: Statistical Society of Slovenia, 2024, S. 21-21. ISBN 978-961-94283-4-4. Dostupné z: https://as.mf.uni-lj.si/uploads/pdf/as2024book.pdf
ID: 647005
Online record: View online record

Abstract

ZÁKLADNÍ ÚDAJE: International Conference APPLIED STATISTICS 2024 Program and Abstracts. Ljubljana: Statistical Society of Slovenia, 2024 - (Kastrin, A., Lusa, L.). ISBN 978-961-94283-4-4. [Applied Statistics 2025 /21./. 21.09.2025-23.09.2025, Koper]. ABSTRAKT: We will describe a flexible semiparametric approach to sequential modeling educational performance of individual students during exam period. In particular, we use complexity penalized splines motivated from expert considerations. The course of a student performance for several mathematically-oriented subjects is modelled via conditional probability decomposition following the ordering given by prerequisites utilizing both the information about the past performance and background individual-specific but time-invariant variables. We show two possible strategies to model identification based on frequentist and Bayesian paradigms. The modular nature of the model allows not only to construct predictions but also to get formalized insights into how the student performance gap propagates. Then we illustrate their performance on real data obtained from a current TACR (Technological Agency of the Czech Republic) project focused on development of an online dynamic system for student support.