Section 4: General and background information
3.6.1.6 Statistical tests used
Once the questionnaires were received, the data analysis started. The researcher followed a systematic process that began with preparing the data for computer entry, that is, by checking, editing and coding the data, followed by entering the data in the computer, and then by data processing and analysis. In analysing the data, the statistical package for social sciences (SPSS) version 12 was used. It is one o f the most popular statistical packages, can perform highly complex data manipulation and
analysis with simple instructions, has a vast number o f statistical and mathematical functions and a very flexible data handling capability, and it can read data in almost any format (Punch 1998).
However, before the analysis began, the type o f data collected was well considered by the researcher. In general, data can be measured on four different scales: (1) nominal;
(2) ordinal; (3) interval; and (4) ratio. Nominal data imply no more than a labelling o f different categories for which there is no meaningful ordering or ranking (Bowerman and O ’Connell 2007; Easterby-Smith et al. 1991). Ordinal data can be ordered or ranked but the distance between the categories is unknown. Interval data can not only be ordered or ranked but the distance between the categories is precisely defined as well; however, there is no inherently defined zero value i.e. the zero point is arbitrary.
Finally, ratio data have all the characteristics o f interval data, and in addition, they have a meaningful zero point (Bowerman and O ’Connell 2007; Fielding and Gilbert 2000; Siegel and Castellan 1988). Saunders et a l (2007) argued that these different types o f data dictate the range o f the techniques available to the researcher for presentation, summary and analysis o f the data collected.
In general, two main statistical techniques have been used in the current study, namely, descriptive statistics and inferential statistics. Descriptive statistics is the first step in the analysis o f data. It entails the researcher in summarising and organising data in an effective and meaningful way. It keeps the researcher close to the data, at least in the initial stages o f the analysis, and helps the researcher in understanding the distribution o f each variable across the survey respondents (Frankfort-Nachmias and Nachmias 1992; Punch 1998). Accordingly, the frequency distribution table and cross-tabulation were used in the current study as the first step in the data analysis.
These methods are discussed in details in Chapter 4.
As the analysis progressed beyond the descriptive stage, the researcher applied the tools o f inferential statistics. The function o f inferential statistics is to provide an idea about whether the patterns described in the sample are likely to apply in the population from which the sample is drawn (De Vaus 2002). It involves using data collected from a sample to draw conclusions about a complete population (Hussey and Hussey 1997). In general, there are two major groups o f statistical tests:
parametric and nonparametric tests. The major distinction between these two groups lies in their power and the underlying assumptions about the data to be analysed (Burns 2000; Zikmund 2000).
In using parametric tests, a number o f assumptions about the actual data need to be satisfied: first, the observations must be independent; second, the observations must be drawn from a normally distributed population (bell-shaped distribution); third, the populations must have the same variance. Finally, the variables must have been measured in at least an interval scale, so that it is possible to interpret the results (Siegel and Castellan 1988).
In contrast, a nonparametric test is based on a model that specifies only very general conditions; it neither specifies the normality condition nor requires an interval level o f measurement. Certain assumptions are associated with most nonparametric tests for example the observations are independent and the variable under study has underlying continuity; however, these assumptions are weaker and fewer than those associated with the parametric tests (Frankfort-Nachmias and Nachmias 1992; Siegel and Castellan 1988).
Nonparametric tests have many advantages. They do not make numerous or stringent assumptions about the population from which the sample was drawn. The error caused by assuming a population is normally distributed, and when it is not, it is avoided. Thus, they are considered as distribution-free tests. Some nonparametric tests are appropriate for data measured in an ordinal scale, and others for data measured in a nominal scale. They often test different hypotheses rather than parametric tests. If the sample is small, there may be no alternative to using a nonparametric test unless the nature o f the population distribution is known exactly.
In addition, nonparametric tests are much easier to apply and their interpretation is often more direct than the interpretation o f parametric tests (Siegel and Castellan
1988; Zikmund 2000).
In general, parametric tests are more powerful than nonparametric tests when all the assumptions are met. The meaningfulness o f their results depends on the validity of these assumptions. If there is any doubt about the quality o f data or the underlying
assumptions, then parametric tests may be unreliable and nonparametric tests should be adopted (Siegel and Castellan 1988; Smith 2003).
Based on the above discussion, and according to the hypotheses o f the current study (Section 3.2.4) and the type o f data collected (nominal and ordinal), nonparametric tests were employed. In particular, the Kruskal-Wallis One-Way Analysis o f Variance, Chi-Square Test o f Independence, and Spearman’s Rank Correlation were used in the current study to analyse the data o f the questionnaire. In addition to these non-parametric tests, multiple regression analysis and particularly stepwise regression were used to investigate the effect o f AIS security controls on the different types of AIS security threats facing UK companies. A full discussion o f these tests and the questionnaire findings is presented in Chapter 4.