Clustering individuals as a way of dealing with multiple predictors in occupational stress research

As the number and type of variables thought to affect the amount of stress experienced by an individual in the workplace increase, and as these effects are often found to be nonlinear and to interact with each other, the forms of statistical analyses employed present different costs and benefits. The use of cluster analysis as a preliminary means of categorizing individuals is described as the least artificial and empirically most accurate means of deriving groups that can lead to hypotheses and hypothesis testing using the more conventional means of analysis of variance. Such a methodology is described as it was applied to an investigation of the effects of demands, attitudes to demands, supports-constraints, and trait neuroticism upon job satisfaction and scores on the General Health Questionnaire for a sample of psychiatric nurses.