Analyzing Hierarchical and Panel Data

MA Social Research module 

Module Outline

Respondents on this course are often clustered within groups – such as children within school classes, people within families (and neighbourhoods), employees within firms.  As such the lowest level units are not independent of each other, and alternative statistical approaches that take account of these ‘hierarchical structures’ should be preferred by researchers. 

One of the most important cases is repeated observations on the same units (panel data, a form of longitudinal data).

This course:

  • Discusses the best way of dealing with data structures in these circumstances – the kinds of data management challenges faced, and how to deal with them in standard statistical software.
  • Covers simple solutions for exploring results from such data (such as using cluster averages; dummy variable & other approaches with a small number of clusters; exploratory approaches using ANOVA).
  • Introduces statistical models that are specially designed to model hierarchical or clustered/structured data which are sometimes known as ‘multilevel models’ (variously known as hierarchical linear models, mixed models and variance components models), and including fixed-effects and random-effects approaches.

Learning Outcomes

At the end of the course, students will be familiar a range of data structures and data types that can be handled by multilevel models.  They will be familiar with a range of exploratory methods of describing the data, and different ways of handling data at more than one level in practical terms.  The course aims to be practical rather than technical, but knowledge at the level of linear regression is assumed, as is familiarity with programming SPSS using syntax.

The course will include how different statistical packages may be used to analyse data of this kind, with some exposure to both generic packages (e.g. SPSS or stata) and specialised packages (e.g. HLM).