Research Flow

29

CHAPTER 3

METHODS AND DATA COLLECTION

30 3.2 Methods

3.2.1 Data Collection

The data used in this research consist of two categories, namely

1. Primary data collected directly by researchers for through direct effort and direct observation in the field and data gathered as needed that it matches the research objectives. Kind primary data gathered in this research are:

• Interview data, and

• Questionnaire data.

2. Secondary data is data that has been collected or produced by another institution / institution / person. More specifically, secondary data is first-hand information that has been collected, recorded, and processed by agencies / institutions / other people with specific objectives. Kind secondary data gathered in this research are:

• Statistical report

• report, book, journal, and

• trusted website

The stage of data collection mainly divided into two stage:

1. Preliminary survey conducted done to get an overview of real conditions in the field regarding the following:

• problems that occur in the field

• research questionnaire testing

• testing methods for conducting surveys

• provide input or improvements that can be used for the main survey

2. Main survey done to get answers to the objectives of the study with considering the results of the preliminary survey so that the survey is on target and runs effectively.

3.2.2 Analysis Methods

The analysis carried out in this study is divided into several types according to the objectives to be achieved in each chapter. in this study there are three chapters which have analysis, namely chapter two concerning road development strategies, chapter six concerning

31

aggressive driving behaviour (ADB) and chapter seven regarding local resident acceptance model. Type of analysis conducted each chapter described as follow:

1. Chapter Two, road development strategies

This chapter using SWOT (Strengths, Weaknesses, Opportunities, and Threats) analysis. SWOT analysis is used to identify and evaluate the strengths, weaknesses, opportunities and threats. These factors analysed using matrix relation strength-opportunity (S-O), weakness-strength-opportunity (W-O), strength-threat (S-T), and weakness-threat (W-T).

2. Chapter Six, Aggressive driving behaviour

Stages of analysis conducted in this chapter are generated using IBM SPSS Statistic 25 as statistical tool, to define the aggressive driving behaviour factor of driving tourists. Those stages are as follows:

• Aggressive driving value analysis

This analysis is conducted to examine and compare the values of aggressive driving generated by aggressiveness (agr) variable that represents the severity of driving behaviour, frequency (fre) that represents the likelihood the driving behaviour occurrence, and weight (wei) that represents the importance of driving behaviour to overall aggressive driving. In this study, aggressive driving is referred to as Aggressive Driving Behaviour Recognition (ADBR) and value of which is based on local resident observations.

The reliability and validity of analysis are evaluated to compare aggressive driving value of the three tests. The first test assumes aggressiveness (agr) as aggressive driving value. The second test assumed that aggressive driving is constructed by aggressiveness (agr), and frequency (fre) is calculated using equation (2), employing approach to risk analysis of equation (1). The third test assumed that aggressive driving is constructed by aggressiveness (agr), frequency (fre), and weight (wei) calculated in equation (3) using simple multiplication process. Aggressive driving value for the three tests are set between 1 to 5, so that those values are comparable. The second test, using the approach to risk assessment (Anghel, 2014), is calculated as follow:

32

𝑃𝑜𝑡𝑒𝑛𝑡𝑖𝑎𝑙 𝑅𝑖𝑠𝑘 = 𝑆𝑒𝑣𝑒𝑟𝑖𝑡𝑦 × 𝐹𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦 (1)

(2) 𝐴𝐷𝐵_𝑅 = √∏(𝑎𝑔𝑟 × 𝑓𝑟𝑒)

𝑗

𝑖=1 2

The third test, applying the approach to risk potential and considered importance (wei) variables, is calculated as follow:

(3)

𝐴𝐷𝐵_𝑅 = √∏(𝑎𝑔𝑟 × 𝑓𝑟𝑒 × 𝑤𝑒𝑖)

𝑗

𝑖=1 3

• Factor analysis

In this study, the type of factor analysis employed is Exploratory Factor Analysis (EFA). EFA method is used to generate a smaller number of latent variables which represent original variables (Henson and Robert, 2006). Reliability and validity analysis should be done before factor analysis is conducted (Wang et al, 2018). Reliability test measures Cronbach's α coefficient and it has value above 0.70 (Hair et al, 2014), validity analysis measures KMO (Kaiser-Meyer-Olkin) whose value is closer to 1, and Bartlett spherical test measures F value sig. <

0.05.

3. Chapter Seven, Local resident acceptance model

Analysis in this chapter are generated using IBM SPSS Statistic 25 and Amos 25 as statistical tool, to define basic model of local resident acceptance and model of local acceptance influenced by aggressive driving. Stage of analysis in this chapter are as follow:

• Descriptive statistical analysis

• Factor analysis, EFA and CFA

Confirmatory Factor Analysis (CFA) conducted to clarify the validity of interrelationships between the factors and confirmed whether each construct suit with proposed model. The values larger than 0.5 indicates that the factors could

33

support each construct (Awang, 2015). The fitness of proposed model should meet the model fit and their level of acceptance as show at table 3-1.

Table 3-1 Model fit and their level of acceptance Name of

Category

Name of

Index Index Full Name Level of

Acceptance Literature

Absolut fit

Chi-Square Discrepancy Chi Square

P-value > 0.05.

not applicable for large sample size (more than 200)

Wheaton et al.

(1977),

RMSEA

Root Mean Square of Error Approximation

RMSEA < 0.08 Browne and Cudeck (1993)

GFI Goodness of Fit

Index GFI > 0.90 Joreskog and Sorbom (1984)

Incremental fit

AGFI Adjusted

Goodness of Fit AGFI > 0.90 Tanaka and Huba (1985)

CFI Comparative Fit

Index CFI > 0.90 Bentler (1990)

TLI Tucker-Lewis

Index TLI > 0.90 Bentler and Bonett (1980)

Parsimonious fit

NFI Normed Fit Index NFI > 0.90 Bollen (1989b)

Chisq/df

Chi Square/

Degrees of Freedom

Chi-Square/ df <

3.0

Marsh and Hocevar (1985),

• Structural Equation Modelling (SEM)

Structural equation modelling is a multivariate statistical analysis technique that is used to analyse structural relationships. This technique is the combination of factor analysis and multiple regression analysis, and it is used to analyse the structural relationship between measured variables and latent constructs. Usually used to testing of a relative set of relationships complicated in stages or simultaneous.

34

Some of the reasons for using SEM analysis are as follows:

• The model analysed is multilevel and relatively complicated, so it will be very difficult to solve the path analysis method in linear regression.

• Able to test complex and multilevel hypotheses simultaneously.

• Errors in each observation are not ignored but are still analysed, so SEM is more accurate for analysing questionnaire data that involves perception.

• Able to analyse the recursive model simultaneously, where this model cannot be solved by linear regression analysis simultaneously.

• There is a bootstrapping facility, where it cannot be done by linear regression analysis.

• Researchers can easily modify the model with a second order to improve the model that has been compiled to be more statistically feasible

SEM includes measurement model and the structural model. In measurement model reliability measure by Cronbach’s Alpha more than 0.7 and validity checked to indicate the model is acceptable. Convergent validity and discriminant validity of the scale were assessed for each latent variable; the CRs were above 0.7 and the AVEs were > 0.5. (Ahmad et al. 2016; Bagozzi and Yi 1988; Fornell and Larcker 1981)

3.3 Data Collection