Researchers

Dr Peter W. F. Smith, Dr Ann M. Berrington, University of Southampton
Dr Patrick Sturgis, University of Surrey

1 June 2003 - 31 May 2005

Context

The project will apply cutting-edge statistical techniques to attitudinal panel data to illustrate the state of the art in how these complex data structures should be approached analytically. The project examines the antecedents and consequences of attitude change, incorporating corrections for measurement error, autocorrelation and nonresponse.

Aims and Objectives

• To investigate the association between life course events such as cohabitation, marriage and divorce and gender/family attitudes;
• To examine the causes and consequences of change in party support and core political values;
• To provide illustrative strategies for choosing the most appropriate methods for analysing repeated measures attitude data;
• To disseminate the results via conference papers, journal articles and short courses.

Methodological Dimensions

• Evaluating stability and change in the social and political beliefs of the British public during the 1990s;
• Examine recursive relationships between demographic events and attitudes;
• Corrections for measurement error.
• Adjustments for differential unit and item nonresponse;
• Accounting for hierarchical data structures.

Research Strategy

• Secondary analysis of survey data;
• Data sources: British Household Panel Study and British Election Panel Studies;
• Multi-level modeling to account for hierarchical data structures;
• Structural Equation Modeling to examine reciprocal effects between demographic events and social attitudes over the life course;
• Model based imputation of missing data. Integration of modeling approaches across methodological and disciplinary boundaries;
• Where necessary, develop new techniques to meet the aims of substantive analysis.

Outputs

A short course and workshop to train social scientists in analysis of repeated measures data, using project outputs as didactic exemplars. Papers from the project will be available on the web and will be disseminated at conferences.