Modelling Satisfaction with Leisure Activities over Time Using Graphical Chain Models
Maria de Fatima Salgueiro, Peter W.F. Smith
Department of Quantitative Methods and UNIDE, ISCTE-IUL - Lisbon University Institute, Lisbon, Portugal; Southampton Statistical Sciences Research Institute, University of Southampton, Southampton, Hampshire, United Kingdom

In recent years subjective well-being (SWB) has been a subject of great interest by psychologists, economists, social scientists and policy makers, and an area of on-going research. Statistical approaches adopted in the literature to model SWB often consider an overall measure or a derived total score. Ordered probit models and fixed and random effects models have often been used, mostly with cross-sectional data.

Panel survey data allow the investigation of stability and change over time. Recent approaches to modelling longitudinal survey data include structural equation models and graphical chain models (GCMs) - see Berrington et al. (2008). Graphical modelling is a statistical technique explicitly based on the concept of conditional independence: the emphasis is on the assessment of the possible relationships between pairs of variables, controlling for other variables under analysis. The independence structure of the variables in the model is represented in a graph. To model longitudinal data, with a natural ordering between the variables, chain graphs with ordered blocks are required, and GCMs have to be considered - see Cox and Wermuth (1996).

The British Household Panel Survey (BHPS) is a national representative survey conducted in Great Britain, since 1991, on an annual basis. Several SWB measures are available in the BHPS. Since 1996 respondents have been asked to rate their satisfaction levels with several domain dimensions of life satisfaction, namely amount of leisure time, use of leisure time and social life.

The current paper discusses the use of GCM to model longitudinal perceptions of satisfaction with leisure activities, and its determinants. BHPS data from years 2002 to 2005 are used. The sample includes 1970 employees, who were original sample members and fully answered all questions under analysis. The proposed GCM has five blocks. Blocks 2 to 5 include the repeated measures of the satisfaction with leisure activities derived score (computed as a sum of three items). Block 1 includes the 2002 measures of the possible determinants of satisfaction: age, gender, marital status, perceived health status, qualifications, social class, number of children in the household, number of hours worked per week and household income.

The capabilities of the software MIM (Edwards, 2000), available for performing model selection in GCMs, are discussed.


Berrington, A., Hu, Y., Smith, P.W.F. and Sturgis, P. (2008): A graphical chain model for reciprocal relationships between women's gender role attitudes and labour force participation. Journal of the Royal Statistical Society, A, 171(1), 89-108.

Cox, D.R. and Wermuth, N. (1996): Multivariate dependencies - models, analysis and interpretation. London: Chapman and Hall.

Edwards, D. (2000): Introduction to Graphical Modelling. 2nd ed. New York: Springer-Verlag.

Keywords: BHPS; Conditional independence; MIM; Panel data

Biography: Maria de Fátima Salgueiro is Associate Professor at ISCTE-IUL - Lisbon University Institute, and member of the Department of Quantitative Methods and of the UNIDE research center. She has a PhD in Social Statistics from the University of Southampton, United Kingdom. Her general research interests include statistical models for longitudinal data, namely graphical models, structural equation models and latent growth curve models, with applications in marketing and in the social sciences.