**4. Methodology**

#### **4.1 Research design, context, population and sample size**

A survey approach is used in this study to gather primary information for empirical analysis. Using a nonprobability—convenience sampling strategy, this research used a deductive methodology [66–69]. Additionally, the sampling method adhered to the guidelines for using structural equation modelling (SEM) [66, 67]. For a covariance-based structural equation modelling (CB-SEM) analysis, Iacobucci [68] suggested a sample size of 200 or more [66–68, 70]. The target audience of this research is thus managers and staff members employed by Malaysia's major five telecommunications companies, namely Axiata Group Berhad, Celcom Digi Berhad, Maxis Communication Berhad, Telekom Malaysia Berhad and TIME Dotcom Berhad [71, 72]. Additionally, G-power software was used to determine the necessary sample size of 210 [73]. However, the 250 questionnaires have been distributed by organisations. Each organisation has been given one of the 50 questionnaires. So, we received 230 questionnaires from managers and their subordinates in order to get the needed sample size response of 210 or more.

#### **4.2 Measurement of variables**

Self-reporting on multi-item measures drawn from earlier research was used to assess each variable. A five-point Likert-type scale was used to assess each metric,

*Innovative Behaviour Mediates in the Relationship between Employee Creativity… DOI: http://dx.doi.org/10.5772/intechopen.111861*

with 1 denoting strongly disagree and 5 denoting strongly agree. The reflective constructs—namely, causality direction, interchangeability, covariation and indicator consequences—were applied in the research investigation [74]. Therefore, the research study included first-order reflective components that were modelled as EC, IB and OI (see **Appendix H** and **Figure H1**).

Additionally, 13 questions from prior research were used to quantify employee creativity as a first-order construct [15]. Innovative behaviour: six questions that examined reflective behaviour as a first-order construct were used [13, 27]. Organisational innovation: it was evaluated using four questions that were accepted as first-order constructs to assess administrative innovation [75].

#### **4.3 Demographic information**

Demographics information (i.e., gender, age, qualification, experience and telecommunications industries) were included in the data. In all, 47% of respondents were male and 53% were female. 21.3% of respondents were less than 29 years old, while 53% were between the ages of 30 and 39 years, 14.8% between the ages of 40 and 49 years, and 10.9% were older than 50 years. In terms of education, 13% had finished high school and secondary school, 26% had a diploma, 52.2% had a bachelor, 7.9% had a master and 0.9% had a PhD degree. In terms of experience, 11.3 per cent had under 4 years, 19.5 per cent had between 5 and 8, 34.8 per cent had between 9 and 12, 15.7 per cent had between 13 and 16, and 18.7 per cent had above 17 years. We obtained 20.4 per cent of our data from Axiata Group Berhad, 20 per cent from Maxis Communication Berhad, 20 per cent from Celcom Digi Berhad, 20.9 per cent from Telekom Malaysia Berhad and 18.7 per cent from TIME Dotcom Berhad about the telecommunications industry. **Table 1** provides the demographics information.

#### **4.4 Data analysis and results**

In order to perform SEM to test the hypotheses, we utilised the Smart PLS 3.2.8 software [66, 67, 76]. This is a reliable and thorough statistical approach that may be used for first-order causal research and does not need making rigid assumptions about the reflective variables [66–74, 76]. The CB-SEM analysis was used to produce bootstrap t-statistics to evaluate the statistical significance of the route coefficients [66–74, 76].

#### **4.5 Measurement model evaluation**

The indicators for each reflective latent variable's individual reliability, construct reliability and convergent validity are provided in **Appendixes C** and **H** (see **Table C1** and **Figure H1**). In addition, indices are provided to help with the precise computation of first-order reflective constructions (see **Appendixes C**, **E**, **G**, **H** and **Tables C1**, **E1**, **G1**, and **Figure H1**). Because their standardised loadings are above the lowest acceptable value of 0.7 [66, 67], the dependability of individual items covering the reflective constructs of EC, IB and OI was judged satisfactory.

Additionally, all of the reflective constructs had composite reliabilities (CR) of 0.7 or above, supporting the construct dependability [66, 67]. Last but not least, the average extracted variance (AVE) was higher than 0.50, confirming convergent validity [66, 67].

Variables showed very little collinearity (see **Appendix G** and **Table G1**), since their individual variance inflation factors (VIF) varied much below the standard cutoff value of 5 [66–74, 76]. Therefore, it may be said that no concept experiences

