*4.2. Measurement*

Entrepreneurial orientation dimensions that are employed in this research were adapted from Awang et al. [28] study. Awang et al. study adapted these items from Lumpkin and Des's [2] study. The dimensions were risk-taking, proactiveness, autonomy, and innovativeness with a total of 12 items. Five-point Likert scale ranging from 1 (strongly disagree) to 5 (strongly agree) was mentioned in the questionnaire. Achievement orientation was measured with scales from Kohli et al. [81] and Chien and Hung [82]. All items were modified for this research. All respondents were needed to agree or disagree with the statements related to an entrepreneurial orientation that best describes their reaction.

Most researchers prefer subjective measures to examine business performance for SMEs. Researchers suggested that subjective measures can be an effective way to examine the business performance of SMEs [83]. Managers of SMEs are always very reluctant to provide financial data to outsiders. The subjective method is used as it is supported by other researchers [84–86]. Out of seven items, a total of six items of business performance ranging from 1 (substantially decrease) and 5 (substantially increase) were adopted and modified for this research from Schalk's [87] study.



#### *4.3. Data Analysis Technique*

To test the hypothetical model this research applied PLS-SEM (partial least square structural equation modeling). Ringle et al. [88] suggested assessing causal relationships through a path model. The PLS path modeling technique is considered one of the general techniques measured by different indicators. According to Compeau et al. [89], both PLS and principal component factors analysis use the component base technique. According to Henseler et al. [80], researchers in management systems, strategic management, and marketing areas used this PLS technique. Research in e-bidding used the PLS-SEM technique to examine the willingness [90]. Aibinu et al. [91], also used the PLS-SEM approach in developing organizational justice modeling and another study assessed the causal relationship between cost overrun and construction resources. In entrepreneurship researchers such as Ferreira et al. [92] also applied the PLS-SEM technique. In HRM practices Triguero-Sanchez et al. [93] study used this PLS-SEM approach. To test the path modeling this research applied Smart-PLS3.0 statistical software. The main intention to use the Smart PLS technique in this research is to identify construct validity and finally assess a path model.

#### 4.3.1. Measuring Construct Validity and Discriminant Validity

In this research construct validity was considered to test internal consistency. To examine the construct validity of the restrained construct's composite reliability (CR) score, average variance extracted (AVE) and Cronbach's alpha tests were used. Discriminant validity and convergent validity results can be seen in Table 1. According to Fornell and Larcker [94], the AVE score should be higher than 0.5. In this research, we can see the AVE values are higher than 0.5, which indicates met the criteria of the acceptable range of convergent validity [80,94,95]. The off-diagonal value of the √AVE is greater than the squared correlation with other constructs, which ultimately met the adequate standard of discriminant validity [80,94].



In Table bold elements, the square root of AVE.

Additionally, the HTMT value was found to be superior to Fornell–Larcker in a variety of scenarios [96]. If the HTMT value is more than 0.85/0.90, it has a discriminant validity problem [96]. The findings of this study are below the 0.90 criterion (Table 2). According to these investigations, the validity of the data appears to be satisfactory.

**Table 2.** Heterotrait–Monotrait (HTMT).


### 4.3.2. Reliability

According to Zhang et al. [97], for assessing survey scales and instruments the most common method used is internal consistency reliability. Cronbach's alpha and composite reliability (CR) are the most common methods used to assess reliability. The CR value can be varied between 0 and 1. In any model, the CR value should be higher than 0.7 [98]. Nunnally [99] also suggested that Cronbach's alpha value should be equal to 0.7 or higher. Wong and Cheung [100] argued that if the value is higher than 0.7, the data is fall into the highly acceptable range. Based on the above discussion it can be summarized that the Cronbach's alpha and CR value should be higher than 0.7. Table 1 shows that all Cronbach's alpha and CR values are higher than 0.7, which indicates acceptable evidence of sufficient convergent validity and reliability.
