Магистратура
2025/2026
Многоуровневые модели
Статус:
Курс по выбору (Аналитика данных и прикладная статистика / Data Analytics and Social Statistics)
Когда читается:
2-й курс, 3 модуль
Охват аудитории:
для своего кампуса
Язык:
английский
Course Syllabus
Abstract
Many data structures are nested: students nested within classrooms, workers nested within business units, observations nested within individuals, et cetera. Until recently, dealing with nested data structures has been difficult both conceptually and computationally. New models that have been termed multilevel models (also known as hierarchical [non]linear models, mixed effects models, or random coefficient models) lead to separating the lower level effects and the higher level effects explicitly into different parts (e.g., Level 1, Level 2, etc.) of the same overarching model. Such models are designed to avoid “aggregation bias” and to solve the “unit of analysis” problem, all while appropriately accounting for the correlated nature of the “within unit” observations. This course will introduce students to the general multilevel model with an emphasis on applications. We will discuss how such models are conceptualized, the meaning and interpretation of the parameter estimates, and finally how to implement them in computer programs. A major emphasis throughout the course will be on how to choose the appropriate model so that specific questions of interest can be addressed in a methodologically sound way.