*STEP 2:*

In this step the model fitting results for Figure 7 were reviewed in light of the selected modification indices (Table 3) from Step 1. The authors considered only those observed variables to be covaried that corresponded to the MI that were equal to/greater than 11.

Covarying can be seen in Figure 8 as small curved two-way arrows between 'e' (error variance) of observed variables in the same factor that were equal to or more than 11. All of the errors (e) were from 'Behaviour', and they represented covariates or correlations between pairs of variables. For example, the MI between 'e13 and 'e14 in Table 3 is 75.421; therefore, it is above the threshold, which was set at 10, and they should be covaried, as shown in Figure 8.

**Figure 8.** Conceptual measurement model of interrelations among building occupants environmental AKB after Step 2.

Figure 8 is the modified measurement model after Step 2, covarying for all the 'e' that are shown in Table 3, which is not ye<sup>t</sup> the best model fit, due to the presence of the low factor loadings shown between each latent variable and their observed variables. There is still some factor loading below 0.1 shown by the arrows in Figure 8. For example, the arrow going from the latent variable 'Behaviour' to its observed variable 'BEH wat 4 was very low at −0.12, and the arrow to 'BEH wat 3 was 0.02.
