Swiss Summer School 2000

Brendan Halpin
An introduction to panel data analysis

Brendan Halpin

Chief Research Officer at the Institute for Social and Economic Research, Essex. Trained as a sociologist, with a DPhil from Nuffield College, Oxford. Research interests in stratification and social mobility, educational homogamy, computer simulation of social processes, methods for the analysis of longitudinal data.

Workshop contents and objectives

This course offers a practical introduction to the manipulation and analysis of panel data. The first aspect concentrates on methods for handling panel data: by its nature panel data is more complex than cross-sectional surveys, and techniques such as aggregation and file-matching are needed to make it amenable to analysis. Techniques covered include moving between household level and individual level, linking cases across waves, creating longitudinal summaries, and combining continuous histories with panel data.

The second aspect of the course is the analysis of panel data. The focus here is also practical, on how to prepare the data, what sorts of analyses are possible, and what sorts of inferences can be drawn. The emphasis will mostly be on relatively simple analyses, using familiar techniques such as linear or logistic regression adapted for the panel context, but some specialised panel models will also be considered (e.g. fixed and random effects models, discrete-time hazard rate models).

Issues specific to panel data such as attrition and longitudinal weighting will also be considered.

The course will mostly use SPSS (and, subject to demand, Stata). The data used will be, in the main, the British Household Panel Study, which is a fairly typical and reasonably mature panel study (eight waves are currently available). Reference will be made to other data sets, such as the German Socio-ecomonic Panel (GSOEP), the Swiss Panel, the European Community Household Panel (ECHP), and the multi-national Panel Analysis Comparability data set (PACO).

The orientation of the course will be practical, with real-life research questions being worked through in the classes and practicals. Students will be encouraged to formulate their own research questions.

Bibliography

Saunders and Brynin, `Ordinary least squares and logistic regression analysis', in Scarbrough and Tanenbaum, Research Strategies in the Social Sciences, OUP 1998.

Prerequisites

A basic grounding in statistical inference and modelling. Some experience with the analysis of cross-sectional survey data. Familiarity with at least linear regression. Familiarity with additional methods such as logistic, probit, loglinear would be helpful.

Basic competence in at least one survey data analysis package such as SPSS, SAS, Stata, etc. Teaching will depend mostly on SPSS.

 

[Back] [Workshop Programme]
EH