4.4.2. Reliability Analysis

Internal consistency and reliability of the dataset was checked using both Cronbatch's alpha values and composite reliability. The alpha values were calculated using SPSS 20, while composite reliability measures were obtained through CFA output. The overall scale provided the alpha of 0.911 while the alpha values were between 0.903 and 0.973 and composite reliability values were between 0.907–0.975 for each latent construct. The Cronbatch's alpha and composite reliability values for each latent construct presented in the model are given in Table 1.

#### 4.4.3. Validity Analysis

Convergent validity can be achieved by getting the loadings of observed variables on their respective latent constructs significant (*p* < 0.001) and the squared multiple correlation value of each observed variable greater than 0.5. The validity analysis results indicated that the dataset was valid for further analysis. The values of squared multiple correlation are presented in Table 2.


**Table 2.** Descriptive statistics and correlations.

\* correlation significant at 0.05. \*\* correlation significant at 0.01 shared variance are in parenthesis. AVE is on diagonal.

Similarly, the discriminant validity for the dataset was evaluated using the criteria presented by Fornell and Larker [78], where the shared variance of any construct should not be greater than the average variance extracted (AVE). The AVE value for every variable was greater than the shared variance of all variables, hence indicating a discriminant validity of the data.

#### *4.5. Hypotheses Testing*

All proposed hypotheses were tested using PROCESS Macro by Hayes [79]. PRO-CESS Macro was preferred over other analytical techniques because of its robustness and bootstrapping approach. The PROCESS Macro provides biased corrected 95% CI and can simultaneously analyze the moderation and mediation effect for complex models.

We have used an incremental approach to test our hypotheses, where at the first step we assessed two mediation models by taking safety consciousness and safety climate as mediators. After that, two moderation models taking safety consciousness and safety climate, respectively, were analyzed. Finally, the full mediation moderation model was assessed. In total, five models using PROCESS Macro have been analyzed.
