As a result, from the pre-test, the ambiguity of some items has been detected. Consequently, the measurements been modified by adding some more items and rephrasing some questions. The final modified questionnaire after pre-testing was used for collecting the data.
22.214.171.124 Administration of Final Questionnaire
The final draft of the questionnaires was administered directly to the target sample of the study. One hundred and sixty (160) copies of the questionnaire have been distributed to respondents and later (145) questionnaires were retrieved with response rate of (91%).
4.8 Data Analysis Techniques
For the purpose of analyzing the data and testing hypotheses, several statistical techniques were applied using (PLS-SEM) through a computer-based tool namely, (Smart PLS 3.0). This study utilizes PLS Model (Smart PLS) rather than CB-SEM Model (AMOS) for several reasons. First, according to Kline (2011) a typical sample size in studies where SEM is used is about 200 cases, whereas the sample size in this study is 160 cases. Second, in a situation where a complex model exist, PLS models can be suitable with limited sample sizes, while some CB-SEM models might not fit. Finally, one of the main advantages of PLS-SEM over CB-SEM is that PLS-SEM can handle numerous independent variables at the same time, even when these display multicollinearity (Hair, Ringle, & Sarstedt, 2011). Moreover, the PLS-SEM approach is useful when it comes to predictions and explanations of target constructs (Hair et al. 2014). Accordingly, this study uses Smart-PLS 3.0 software as well as SPSS 24.0 particularly for data examination and descriptive statistics.
The following subsections presents numerous statistical techniques used to analyze the survey data as follows:
4.8.1 Descriptive Statistics
For the purpose of describing both the responding firms and respondents, frequency and percentage were used. As well as, this study uses descriptive statistic, namely means, standard deviation, maximum and minimum to describe the variables of study.
4.8.2 Evaluating Measurement Model
This study uses reflective measurement model. Therefore, according to prior studies, the validation of a reflective measurement model can be established by testing its internal consistency, indicator reliability, convergent validity and discriminant validity (Lewis, Templeton, & Byrd, 2005)
126.96.36.199 Internal Consistency
To evaluate internal consistency this study uses Composite Reliability (CR), although traditionally, internal consistency is evaluated through Cronbach’s Alph(CA). Nonetheless, in PLS, internal consistency is measured using CR. Owing that CR considers that items have different loading. However, this study uses both CR and CA. Since, CA also provides somehow accurate estimation of internal consistency, therefore, CR will be used as higher pound and CA as lower pound. For establishment of internal consistency, composite reliability should be higher than 0.70 (Hair et al.2017)
188.8.131.52 Indicator reliability
The purpose of measuring Indicator reliability is to evaluate the extent to which a variable is consistent with what it intends to measure. The outer loadings of item should be greater than 0.70. items with outer loadings between 0.40 and 0.70 should be considered for removal only if the deletion leads to an increase in composite reliability (Hair et al.2017).
184.108.40.206 Convergent Validity
Convergent validity takes two measures that are supposed to be measuring the same variable and shows that they are related in measuring same variable. (Urbach & Ahleman, 2010). In PLS convergent validity can be evaluated using the value of average variance extracted (AVE). According to Hair et al.2017 sufficient convergent validity is achieved when the average variance extracted ( AVE) value of a variable higher than 0.5.
220.127.116.11 Discriminant Validity
Discriminant validity or divergent validity tests whether measurements that are not supposed to be related are actually unrelated. discriminant validity evaluates whether indicators do not measure something else (Urbach & Ahlemann, 2010). According to Henseler et at., (2015 to measure discriminant validity in PLS, heterotrait-monotrait ratio (HTMT) of the correlations should be examined (Hair et al.2017). HTMT is measured as criterion value which has to be lower than (0.85). As well as HTMT is measured as a statistical test and in this case HTMT inference should be with upper internal confidence (<1). Fulfilment of these criterions result in establishment of discriminant validity. (Henseler et al., 2015).
4.8.3 Correlation Analysis
Correlation measures the strength and direction of a linear relationship between two variables. In addition, correlation is used to detect collinearity among variables of study. The values of correlation coefficient vary from –1 to +1. Exactly (+1) indicates a perfect positive linear relationship and exactly (-) indicate negative relationship, Whereas, (0) Indicates no linear relationship. Correlation of 0.30, 0.50 and 0.70 indicate weak, moderate and strong relationship between variables, respectively. (Clarke-Pearson, et. al.”,1988)
4.8.4 Evaluating Structural Model
This phase of analysis is conducted after confirming the validity and reliability of measurements. The aim of evaluating structural model is to test the proposed hypotheses.
Five stages are followed to assess the structural model, namely collinearity assessment, significance of the path coefficients, the level of the R2 values, the f2 effect size and the predictive relevance Q2 (Hair”,2017). These statistical techniques are discussed in the following chapter.
4.8.4 Multigroup Analysis
Multigroup is a type of moderator analysis where the moderator variable is categorical and is anticipated to potentially affect the relationships between the vocal variable of study. Therefore, Multigroup analysis allows testing whether differences between group-specific path coefficients are statistically significant. (Hair”,2017). This study proposes several control variables more precisely, Industry type, firm size, firm age, firm ownership, firm ownership and number of competitor. Consequently, Multigroup analysis is used in this study test the differences between these groups and their influence on the vocal variables of study.
4.9 Research Ethical Considerations
Conducing a research requires not only expertise and diligence, but also honesty and integrity. This is done to recognize and protect the rights of human subjects. Therefore, there has to be some basic ethics to be adopted in any research.
Consequently, in this study the researcher adheres to ethics by keeping the answers acquired strictly confidential. furthermore, a prior permission was taken from the target respondents. Moreover, the researcher acknowledge the previous literature work through making references and citations to any quotation, paraphrasing and summary extracted from others’ work.
Since, scientific honesty is regarded as an important ethical responsibility when conducting research, dishonest conduct may include manipulation of data (Brink 1996). Therefore, the researcher tried to avoid any form of dishonesty by providing the dataset to an independent statistician who independently analyzes the data and produces the results to avoid any subjective collaboration.