Fig 1.
The research framework for investigating the stakeholder engagement variability.
Fig 2.
An illustration of a simple network with five nodes and five edges.
Fig 3.
Three prototypical network structures.
Fig 4.
Edge thick in the aggregated network proportionates to the edge weight.
Fig 5.
Microstructures or configurations of the exponential random graph model are considered in this study.
Table 1.
Basic network statistics of the 56 stakeholder networks resulting from the survey response data used for this study.
Table 2.
Top-5 nodes that appeared most in the three different types of stakeholder networks.
The proportion of percentage appearance in parentheses follows the count value. All values in this table are significant at p<0.001, resulting from a t-test.
Table 3.
Results from the one-way analysis of variance (ANOVA) test for the various network measures across three different project categories.
Fig 6.
The kernel density estimations of four network measures.
Table 4.
Results from the independent sample t-test for four network measures.
Each project is grouped based on its cost performance (within budget or not).
Table 5.
Top-5 nodes or stakeholders in the three different aggregated networks.
Fig 7.
Visualizations of the three aggregated networks: (a) public, (b) private, and (c) public-private partnership. The size of a node proportionates to its degree centrality value. The thickness of an edge proportionates to its edge weight.
Table 6.
The estimation results from the exponential random graph model: (a) aggregated public network; and (b) the aggregated private network.
The aggregated PPP network does not converge. AT stands for ’alternative-k-triangles’.
Table 7.
Common stakeholders among the four top-5 lists (for the public, private and public-private partnership project categories) from Tables 2 and 5(A)–5(C).
We follow the alphabetical order in placing them on the table.