Skip to content

AngelGarciaODiana/Master-Thesis

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

10 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Master-Thesis DOI

The repository have all supplementary materials for the Master's thesis submitted to the Universidad César Vallejo, as part of the University Teaching Master Program. Please feel free to use any of these files as part of your research project, only give me the correspondent attributions in your citations.

Dissertation slides can be found on quarto repository:
QUARTO
If you use this slides in your work, please cite it using the following metadata:
DOI

Análisis Comparativo de Técnicas de Estimación en el Análisis de Redes Aplicado a la Investigación en Docencia Universitaria

Comparative Analysis of Estimation Techniques in Network Analysis Applied to Educative Research


El objetivo del estudio fue examinar el desempeño de cuatro métodos de estimación de redes (EBICglasso, huge, TMFG, LoGo) y tres algoritmos de detección de comunidades (walktrap, leiden, spinglass) en bases de datos de investigación educativa. Empleó un diseño experimental basado en datos con simulaciones Monte Carlo. A partir de dos conjuntos de datos provenientes de escalas ordinales tipo Likert con características compartidas, se generaron 988,800 casos a través de diversos tamaños de muestra. Finalmente, fueron un total de 192 modelos estimados a través cuatro métodos de estimación de redes bajo dos estrategias de conversión de datos (normal y no paranormal). Se realizó un análisis de clústeres para cada modelo, lo que dio lugar a 576 redes únicas. El análisis final, replicado para dos variables, resultó en 1,152 estimaciones. Las conclusiones indican que una cuidadosa selección de configuraciones mejora la replicabilidad y consistencia, además que las configuraciones iniciales de estimación son cruciales para obtener resultados precisos. El diseño de modelos dinámicos y complejos a través del análisis de redes psicométricas puede conducir a nuevos paradigmas pedagógicos y andragógicos, lo que permitirá optimizar los mecanismos educativos y comprender el desarrollo integral de los estudiantes.
The aim of the study was to examine the performance of four network estimation methods (EBICglasso, huge, TMFG, LoGo) and three community detection algorithms (walktrap, leiden, spinglass) in educational research databases. It employed a data-driven experimental design with Monte Carlo simulations. From two datasets derived from Likert-type ordinal scales with shared characteristics, 988,800 cases were generated across various sample sizes. Ultimately, a total of 192 models were estimated using four network estimation methods under two data conversion strategies (normal and non-paranormal). A cluster analysis was performed for each model, resulting in 576 unique networks. The final analysis, replicated for two variables, yielded 1,152 estimates. The findings indicate that a careful selection of configurations improves replicability and consistency, and that the initial estimation settings are crucial for obtaining accurate results. The design of dynamic and complex models through psychometric network analysis can lead to new pedagogical and andragogical paradigms, enabling the optimization of educational mechanisms and a better understanding of students' holistic development.
METADATA: Network Analysis, Cluster Analysis, Simulation, System Comparison, Student Engagement, Mathematical Anxiety, Spinglass, Walktrap, Leiden, Louvain, EBICglasso, TMFG, LoGo, huge, Adjusted Rand Index, Network Comparison Test, Monte Carlo Method.

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages