An investigation into statistical methods for analysing ordered categorical data.

This investigation researches statistical methods for analysing ordered categorical data. Some standard descriptive and modelling procedures are described, and the data is analysed using a relatively new statistical package, CHAID, which is designed purely for categorical data analysis. The study is centered around the application of the proportional odds and continuation odds models, to data obtained from a survey of the opinions of South Yorkshire Police staff (SSRC (1994)). Morale within the South Yorkshire Police is the factor of interest, and is discussed in some detail. The two approaches of proportional odds and continuation odds models are discussed critically. Dummy variables and scored levels are employed for the treatment of ordinal variables. The effects of these two methods of coding ordinal data, on the results of the analyses, are also compared and discussed. Methods of assessing the goodness-of-fit of ordinal models are discussed, and a modification to the guidelines for using a recently presented technique (Lipsitz et al (1996)) is suggested and applied. The proportional odds model is successfully applied. The implications from the models produced are that job satisfaction, communication, public view of the police, promotion issues and length of service have an influence on the morale of an individual, in general.