Multilevel regression analysis | 38/39 | UPV

Título: Multilevel regression analysis


Descripción automática: In this video, the presenter introduces the concept of multilevel regression models, particularly applied to educational data analysis, where traditional regression assumptions, such as the independence of data points, may not hold. The video explains that in educational settings, data are often structured hierarchically, with students nested within schools, classes, or countries, leading to potential correlations among data points within the same groupings.

The presenter discusses how multilevel regression can address research questions that involve this hierarchical data structure. A case study by Elena Meroni and Eduardo Vera is referenced to illustrate the analysis of student performance relative to teacher training quality in reading skills, taking into account the different teaching environments and varying student performance across countries and schools.

Key distinctions between random and fixed classifications in defining datasets are explained, reflecting whether the data represent a sample or the entirety of the population. Additionally, the presenter outlines two mathematical models used in multilevel regression – the random intercept model and a more complex version allowing for varying slopes.

The video underscores the importance of the intraclass correlation coefficient in determining the extent to which variability in a dependent variable, like student performance, is attributable to differences between groups such as schools. It suggests applying multilevel regression models when this coefficient is significant, indicating substantial between-group variability.

Finally, the case study is revisited, detailing how the authors assessed the influence of various factors on student reading skills, distinguishing between random (country, school, student level) and fixed effects (gender, socioeconomic status, etc.), and identifying which had positive or negative influences on student performance. The presentation concludes with a summary of these findings and an invitation to future videos.

Autor/a: Conchado Peiró Andrea

Curso: Este vídeo es el 38/39 del curso MOOC Educational Data Science.    • MOOC Educational Data Science  
Inscríbete en: https://upvx.es/courses/course-v1:Edu...


Universitat Politècnica de València UPV: https://www.upv.es
Más vídeos en:    / valenciaupv  
Accede a nuestros MOOC: https://upvx.es

#regression #predictive models #hierarchical structures