**STAGE III Field survey**

**4.6 Data Analysis**

The last stage of the study attempts to analyze the findings regarding the characteristics and unique features in all case studies selected, and also the established regulations, policies, plans and guidelines impacting conservation and development of small

### University

### of Malaya

140

Malaysia towns against the backdrop of the document reviews. As Stephens (2009) pointed out, data analysis is what the researcher does with data obtained in order to develop explanations of events so that theories and generalizations about the causes, reasons and processes related to the subject study can be developed.

Viewing the multi-strategy research, the data generate in the present study will be analyzed using the Statistical Package for the Social Sciences (SPSS) software, version 17.0 and the NVivo software, version 8.0. Each is used to organize quantitative and qualitative data respectively and will be discussed in the following section.

**4.6.1 Statistical data analysis **

De Vaus (1996) argues that the statistical techniques have to be appropriately matched to the number of variables being examined (univariate, bivariate or multivariate analysis), level of measurement of the variables (nominal, ordinal or interval data) and the purpose for which the data is used (descriptive or inferential). Therefore, the following statistical techniques were implemented using the SPSS software to analyze the survey data in this study.

**4.6.1.1 Frequency distributions and descriptive statistics **

Descriptive statistics was used to describe and summarize patterns of a single variable.

These included respondents‟ socio-demographic characteristics and other variables on the perceptions of the elements and qualities associated with identity of the towns. As argued by De Vaus (1996), these statistics are the most productive in terms of understanding any phenomenon at a fairly early stage of a particular research. The distribution of data are shown in both tables and graphs particularly the histograms as this is the easiest way to describe the basic patterns of numerical data for the variable in question (Neuman, 2006). In summarizing the distributions of the collected data, calculation of central tendency using the mode and mean were also carried out.

### University

### of Malaya

141

**4.6.1.2 Bivariate analysis **

Bivariate analysis deals with analyzing two quantitative variables at a time in order to discover whether the two variables are related. Crosstabulations or contingency tables constitute one of the most useful analytical tools of analyzing association (Bryman, 2004; De Vaus, 1996). Nothwithstanding the fact, they are only appropriate when dealing with variables with less than seven or eight categories each.

Alternatively, the correlation co-efficients or measures of association were used to provide concise summaries of the association in a crosstabulation (Chua, 2006). The coefficient indicates the strength of a relationship and will lie between 0 (no relationship) and 1 (perfect relationship). Nevertheless, it is nearly impossible to have perfect correlation between variables as people never behave in exactly the way the researcher would expect (Chua, 2006; Diamond, 2006). Therefore, the following Table 4.7 will be useful to provide the study with valuable information.

**Table 4.7: Indication of the strength of correlation coefficient (Chua, 2006) **
**Value of correlation coefficient ** **Strength of correlation **

1.00 Perfect relationship

0.91 – 0.99 Very strong

0.71 – 0.90 Strong

0.51 – 0.70 Moderate

0.31 – 0.50 Weak

0.01 – 0.30 Very weak

0 No relationship

For the purpose of this study, two correlation tests were employed as follows:

**a) ** **Chi-square correlation coefficient **

The Chi-square correlation coefficient is a measure of association between two nominal or categorical variables (Bryman, 2004; Chua, 2006; Dancey & Reidy, 2008; De Vaus, 1996). Two chi-square based correlation coefficients used in this study are phi (φ) and

### University

### of Malaya

142

Cramer‟s V. Specifically, the phi coefficient is used for the analysis of the relationship between two dichotomous variables (variables with two categories). On the other hand, Cramer‟s V is recommended for calculating correlation between variables with more than two categories. For example, this coefficient was used to detect significant correlation between gender and ethnicity with the first element noticed in the towns.

**b) ** **Spearman’s rank correlation coefficient **

This bivariate correlation test is used for ordinal level variables (Bryman, 2004, Chua, 2006; Dancey & Reidy, 2008). For example, the Spearman‟s rank correlation coefficient or Spearman‟s rho (represented with the Greek letter p) was used to detect a relationship between ages with the importance of preserving historical resources which developed on a continuum of „strongly disagree‟ to „strongly agree‟. While this study involves exploring relationships between variables with mixed levels of measurement, the selection of the methods of bivariate analysis is also made based on the table provided by Bryman (2004, p.230) and the following approaches by De Vaus (1996):

**Dichotomous variables **

As a rule of thumb, if one variable has only two categories we can ignore its level of measurement and let the other variable determine the choice of the coefficient. For example, if a dichotomous nominal level variable (gender) is crosstabulated with an ordinal level variable (importance of preserving historical resources), both variables can be treated as ordinal and select the appropriate correlation coefficient (Spearman‟s rho)

**Use a weaker statistic **

When neither variable is dichotomous, treat both variables as though they both are at the same level of measurement of the variable measured at the lowest level. For example, if one variable is nominal and the other is ordinal, treat both as they are nominal

### University

### of Malaya

143

**4.6.2 Analysis of qualitative data **

The analysis of qualitative data is usually seen as a somewhat more complex in contrast to quantitative data analysis (Basit, 2003; Casterle et al., 2012; LeCompte, 2000;

Marlow, 2011). The difficulties include the sheer mass of data and the time-consuming nature of qualitative analysis. For Basit (2003) and Casterle et al. (2012), the central difficulty however lies in the absence of a standard procedure for the analysis. For this reason, the qualitative data analysis is characterizes as a dynamic, less formulaic, intuitive and creative process of inductive reasoning, thinking and theorizing.

Consequently, many researchers have encouraged the use of computer tools to facilitate the analysis (Drisko, 1998; Jones, 2007; Miles & Huberman, 1994; Willis & Jost, 1999).

For Blaxter (2010), Blismas & Dainty (2003), Drisko (1998) and Silverman (2005), the software extends the benefits of using computers and allows more thorough analysis by speeding data base management, text segmenting, coding, and theory development task.

For the purpose of this study, qualitative data obtained in the semi-structured interviews were analyzed using the NVivo 8.0 software. Like other qualitative software packages, NVivo is not an analytical tool in the pure sense as it does not replace the necessary and continuous intuitive input of the researcher (Basit, 2003; Blismas & Dainty, 2003;

Casterle, 2012). Notwithstanding the fact, the greatest strengths of NVivo lie in data management, manipulation and most importantly in codification which possesses as the most crucial aspect of qualitative analysis, hence offers a considerable advantage over other methods in this regards (Blismas & Dainty, 2003).

In this study, the interviews are transcribed verbatim before any data coding and analysis. The transcripts are then coded into four category nodes according to major themes outlined earlier in Section 4.5.4. The next level of coding entails the creation of concept nodes whereby each question listed in the interview guide is coded to the next

### University

### of Malaya

144

level. A total of fifteen concepts are coded to the four category nodes. The final level of coding involves the creation of construct nodes which summarizes views of the interviewees. The summary models visualizing the connections between various dimensions of constructs, concepts and categories are shown in Appendix G.

Employment of NVivo software which makes possible the coding of data in terms of the tree nodes is believed to capably deal with qualitative data and hence, assist the overall process of data analysis.