Best practice 66 companies were expected to have

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: n analysis techniques were supported – specifically, if there had been an attempt to introduce corporate definitions of the key terms risk and uncertainty. Best practice 66 companies were expected to have implemented definitions that were complementary to their decision analysis approach. Where there were numerical ties according to these criteria, the tie was broken on the basis of other material from the interviews, indicative of level of sophistication, which was not available on all companies and therefore not included as an overall rank measure (for example, company-wide application of a piece of software). This ranking scheme is discussed further in Chapter 7. Performance measures were then selected that are indicative of business success in the upstream. The choice of these outlined and justified in Chapter 7. The appropriate performance data was then gathered on each company. For some of the criteria, it was only possible to access ordinal level data. For the some however, categorical data was available. The relationship between the rank each company achieved in the decision analysis ranking and their rank, or otherwise, on each performance measure were then analysed together statistically. There is a large number of statistical techniques available for analysing any given set of data. The author has chosen in this thesis to use those tools known as nonparametric or distribution-free. These techniques may be contrasted with others known as parametric techniques. Parametric techniques make a large number of assumptions regarding the nature of the underlying population distribution that are frequently untestable. Leach (1979) argues that social scientists using parametric statistical analysis are taking a gamble. If the population assumptions are correct or approximately correct, then the researcher has very good test. However, if the population assumptions are incorrect, then a non-parametric test may well give a more accurate result. Non-parametric tests make relatively few assumptions about the nature of the data and hence are more widely applicable. Finch and McMaster (2000 p19) write: …non-parametric techniques do not invoke such restrictions. Techniques involved in measuring association do not require the employment of cardinal measures redolent of interval scales. Instead the only measurement requirement is that ordinal scales can be deployed (Lehmann and D’Abrera, 1975; Siegel and Castellan, 1988).” Since only ordinal level data were available for some of the performance criteria and there were no ties in the data, the primary non-parametric technique that the 67 researcher elected to use was Spearman’s rank correlation test. It is outlined in Appendix 3. Spearman’s rank order correlation coefficient is a modified form of the more typically used Pearson’s correlation coefficient. It is mathematically equivalent to Pearson’s correlation coefficient computed on ranks instead of scores. Just as Pearson’s correlation coefficient is interpreted as a measure of linearity, Spearman’s correlation coefficient can be i...
View Full Document

This document was uploaded on 03/30/2014.

Ask a homework question - tutors are online