Quantitative Data Analysis

Quantitative data is analysed using descriptive and inferential statistics. Descriptive statistics helps to define what is or what the data showed directly and used to describe nominal or categorical variables. However, it can also apply to ordinal variables, such as level of agreement in Likert scale responses. (Field, 2013; Pallant, 2013). Descriptive statistical tools include frequency counts and proportions, measures of central tendency and variation (Fink, 2006) which helps in analysing across cases one variable at a time and presented, using tables or graph.

149 On the other hand, inferential statistics seeks to explore group comparisons looking for patterns and relationships in the data with the purpose of drawing conclusions (Dawson, 2009: Field, 2013; Pallant, 2013). It is used to examine relationship or group memberships. To test differences, chi -square test- compare observed frequencies to expected frequencies, T-test compare means of two groups and one-way ANOVA test compare more than two groups (Pallant, 2013). Multiple regression or correlation analysis examines the strength of relationships between variables while exploratory factor analysis or logical group regression is most appropriate for group memberships (Field, 2013: Pallant, 2013). However, the choice of the methods will depend on the nature of the data and what need exploring in the data set (test differences examine relation, group membership or a combination). Owing to the anticipated volume of data, the Statistical Package for the Social Sciences (SPSS 20) software package would be used to facilitate better and clearer analysis.

The questionnaire is structured in both ordinal and categorical scales question format. The questions relating to respondent’s background are expected to produce categorical data while those relating to respondents ratings will produce ordinal data. Part A comprised mainly of categorical scale while part B comprised more of ordinal scales. Therefore, both descriptive and inferential statistics are applicable. The data from the questionnaire will be analysed using the following statistical methods.

3.9.2.1 Frequency counts and proportions

Frequency counts and proportions relate to number counts and percentages. This study adopts percentage proportions to analyse the quantitative data and presents them using tables and bar charts. This analytical tool was considered most appropriate since mean and standard deviation are invalid parameters for descriptive statistics whenever data are on ordinal scales (Gamage, 2011), and it would aid easy understanding of the data results.

3.9.2.2 Relative Agreement Index (RII)

To generate an index of the perceived agreement of responses the Relative agreement index (RII) is used. According to Holt (2014), the RII may be described to reflect contextual application though the generic recognised term is relative importance index and its acronym RII. The RII is used to calculate the strength of index familiarity frequencies and ratings of

150 responses on the frequency of use, the level of influence and level of importance. The RII is calculated using the equation 7.

…… Eqn 7

Where: W– is the weight given to each item by the respondents and ranges from 1 to 4 on the Likert scale,

A – is the highest rating (i.e. 4 in this case) and; N – is the total number of respondents.

3.9.2.3 Kruskal-Wallis Test

The Kruskal-Wallis test is a non-parametric test of analysis of variance (ANOVA) by rank appropriate for the data collected on an ordinal or interval scale that is non-parametric (Pallant, 2013; Cooper & Schindler, 1998). A test compares differences amongst responding groups on an item or group of items in a questionnaire. It is used when the data is not normally distributed (Dawson, 2009; Field, 2013; Pallant, 2013). The Kruskal-Wallis test measures chi-square in its analysis and the degrees of freedom (df) for the chi-square statistic are equal to the number of groups responding to the questionnaire minus one (df-1). Hence, when the value of the asymptotic significance is greater than 0.05, it clearly indicates that there is no difference between the responses from the group respondents by rank (Pallant, 2013). In this research, the Kruskal-Wallis H test was used to establish if the responding groups (HAST, CST, and CNT) have differences in their views concerning a rating question, or variable in the questionnaire.

3.9.2.4 Exploratory factor analysis

Factor analysis FA is a statistical set of techniques applied to reduce a larger set of variables/drivers into a smaller set 'principal components', of” Factors” which account for most of the variance in the original factors (Pallant, 2013). Hence, variables that are correlated and largely independent of other subsets of variables are combined into factors. Various scholars, such as Xu (2013), have employed the FA in their studies to group IMSFs. In this context, the factor analysis (FA) is used to further analyse the results from the RII on implementation

151 drivers. It offers not only the possibility of gaining a clear view of the data but also using the output in subsequent analyses (Pallant, 2013). The study adopts the FA analysis procedure posed by Chan et al., (2004) which involves:

 Identify the drivers

 Compute the correlation matrix for all the drivers;  Extract and rotate driver into factors.

 Interpret and label the factors

A summary of the quantitative data analysis techniques used in this study is summarised in Table 3.11.

Table 3. 11: Data analyses methods employed in Relation to Questionnaire Sections