Content area
Full Text
Llinos Haf Spencer reports on the use of the duster analysis statistical technique in nursing research and uses data from the Welsh Language Awareness in Healthcare Provision in Wales survey as an exemplar. She concludes that cluster analysis is a valuable tool to tease out patterns in data that are not initially evident in bivariate analyses and thus should be considered as a viable option for nursing research
key words
* bilingual
* cluster analysis
* language
* questionnaire survey
* Welsh
Introduction
According to Beckstead (2002), 'cluster analysis is a generic name for a wide variety of methods or procedures that can be used to create a classification. These procedures empirically form discrete groups or "clusters" of similar entities based on similarities among their measured features. Entities may be individuals, objects, or variables.'
The clustering of the variables is undertaken in a way that maximises between-group variation and minimises within-group variation. On this basis, authors such as Hair et al (1992) claim that cluster analysis is the most appropriate statistical technique to delineate natural groupings in data. Furthermore, Blashfield and Aldenerfer (1988) suggest that cluster analysis can be useful to researchers in developing a typology or classification, to investigate useful conceptual schemes for grouping entities, and to generate and test hypotheses through data exploration.
ClustanCraphics (Wishart 2003) is one of the software packages that can be used to undertake cluster analysis. In ClustanGraphics. the mathematical procedure is demonstrated graphically using a branching diagram known as a dendrogram, whereby the number of clusters is evident spatially.
Although seemingly underused in health-related survey studies, cluster analysis has been employed effectively by nursing and healthcare researchers such as Hillhouse and Adler (1997). Beckstead (2002), Thorpe and Loo (2003), Livneh et al (2004) and Chan et al (2004) to group participants in a meaningful way.
Hillhouse and Adler (1997) applied cluster analysis as an exploratory tool for investigating stress effect patterns among hospital staff nurses. Cluster analysis enabled the researchers to demonstrate that stress among nurses is more closely associated with the work environment and overall workload than, as anticipated, with the degree of specialisation in the unit.
Chan et al (2004) adopted cluster analysis to investigate nurses' attitudes towards prenatal bereavement support. They found that a...