Published Papers - Abstract 88

Powers JR & Dobson AJ. Stability of groups of correlated variables identified by exploratory factor analysis and cluster analysis: Example from the Australian Longitudinal Study on Women's Health. Australian and New Zealand Journal of Statistics, ; :

In epidemiological studies, data are commonly collected on a large number of variables. Analyses of these data are complicated by correlations among these variables. Various methods, such as cluster analysis or factor analysis, can be used to identify distinct groups of related variables. The robustness or stability of these groups is important if they are to form the basis for further analysis. In this paper we examine the problem using data obtained from the baseline survey of 14,100 women (aged 45-50 years) enrolled in the Australian Longitudinal Study on Women’s Health (ALSWH). Five methods of identifying groups are used with three different subsets of the data. Those variables which consistently group together could be used to create composite variables, such as factor scores or summed scores, or to identify one variable to represent the group. For the ALSWH data the stable groups of variables represent coherent, easily interpreted aspects of health (physical and mental health, use of health services, gynaecological health, lifestyle and demographic characteristics). This example illustrates the practical issues of reducing multicollinearity.