Questionnaire survey analysis

There are several software packages available for quantitative analysis, such as SAS, STATA, MATLAB, R, and Statistical Package for Social Sciences (SPSS). In this study, data from the questionnaire survey was analysed using SPSS software version 17.0. It was selected due to its relative simplicity compared to other software, and most importantly, it can satisfy the researcher’s data analysing requirements. The analysis was conducted through descriptive and inferential statistics. Descriptive statistics are a method to summarize, describe or display quantitative data (Collis and Hussey, 2009). Descriptive statistics include the measurement of central tendency (mean, median, and mode), measure of dispersion (range, standard deviation, variance, minimum, maximum, and standard errors mean), percentile values (quartiles and percentiles), and measurement of distribution (skewness and kurtosis). In order to represent the particular group of data (i.e. advantages of CPHRP) in a number, this study uses mean as a method of analysis. Further, to analyse the variation of the data, standard deviation analysis is used.

Inferential statistics are used to examine whether the data differs from the hypothetical value, i.e. in CSFs, certain factors are classified as critical if the mean is higher or equal to 4.00. Accordingly, the researcher used t-test to do this. More specifically, since we only compare one group of data to its hypothetical value, one sample t-test is the most appropriate method. The researcher chose the two-tails test to satisfy the confidence level (95%) rather than the one-tail test since the first is more powerful than the latter. The process by which this analysis is carried out is that the researcher decides the minimum value required (i.e. ≥4.00). The researcher then includes this value to be set as population mean. Finally, the researcher analyses whether the factor’s scores are different from the population mean. If they are different and more than the minimum value required, then these factors would be appropriate to use .

Raw data from the questionnaire survey is inputted manually to SPSS software using the specific code (see section 3.9.1.3). For instance, for one particular advantage of the community-based approach, where a respondent ticked ‘extremely important’, the code input to the SPSS software is ‘5’. Having inputted all the raw data, it is analysed to

find out the means and standard deviation of each factor, including the one sample t- test.

For advantages, the variable will be categorised as ‘very significant’ if the mean is equal to or more than four (≥ 4.00). It was based on the categorisation used in the 5- Likert scale in the questionnaire, where 1 is labelled as ‘not significant at all’, 2 as ‘slightly significant’, 3 as significant, and the highest rate 5 as ‘extremely significant’. In cases of more than one factor having the same mean value, the factor with the lower standard deviation is classified as more significant as it implies that the data point tends to be closer to the average compared to the higher standard deviation.

The same conditions are applied in deciding the critical success factors (CSFs). The CSFs question also deployed a 5-Likert scale, starting from 1 (not influential at all) to 5 (extremely influential). To be classified as CSFs, the value of the mean of success factors has to be equal or more than four (≥ 4.00), which indicates that the level of influence has to be more than ‘influential’ – it has to be at the level of very influential or higher. Several studies of CSFs, such as Lu et al. (2008), Shen and Liu (2003), and Kulatunga et al. (2009), also set the level to be classified as CSFs at as four or above. If one factor has the same average, the factor with the lower standard deviation would be classified as more influential, which means that it does not vary or disperse greatly from the average.

Risk assessment is conducted by multiplying the probability of risk occurrence with its impact factor. The analysis is into what extent risk occurrence will affect the project objectives, which are time, cost, quality, and beneficiaries’ satisfaction. For this analysis, the first step is to calculate the average of probability of a particular risk and the average of risk impact. After having the average of risk probability and risk impact, these two numbers are multiplied, and the result is called the probability-impact (PI) factor. For example, the average probability of the risk of ‘unclear reconstruction policy’ is 0.7, and the average of its impact on time is 0.4. Then the PI factor is: 0.7 x 0.4 = 0.28. To determine whether the impact is classified as a high risk or not, the probability impact matrix in Table 2.13 is deployed as reference. This PI matrix is designed by PMI (2008). Moreover, PMI (2008) states that it is dependent on the

organisation to describe the parameters of probability and impact (description of to what extent the probability is low or high, or to what extent the impact is low or high) and also for the classification of high risk, i.e. what PI factor a risk can be classified as high. For this research, risk is classified as ‘high risk’ if the probability impact index is higher or equal to 0.20. The reason behind this number is that it was found that the probability is ranges from 0.5 to 0.7 and impact between 0.2 and 0.4, so looking back at Table 2.13, the researcher decided that the PI factor 0.20 is the minimum value for a risk to be classified as a high-risk event.

Having described all the parameters in the methodological framework, from research philosophy to date collection techniques and analysis, Figure 3.10 summarizes the research methodology to be conducted for this research.

Figure 3.10 The research onion of this study (adapted from Saunders et al., 2009)

Interviews Questionnaires Research Philosophy Research Approach Research Strategies Research Choice Time Horizon Techniques and procedures Pragmatism Inductive- Deductive Case study Mixed method Cross- sectional