4.3.3. Solutions

In this section, we briefly come back to the discussion of the limitations of the extended calibration procedure of Section 3.2. It will be shown that all the limitations are now solved by using a combination of the two options for the selection and stopping strategies. A summary of these solutions is also given in the right column of Figure 3.

First of all, thanks to the implementation of the selection strategy, three of the six limitations have been solved, as follows:


• **Solution 6.** Thanks to the use of the *ei* criterion, 1000 iterations are no longer necessary (S3). Instead, the statistical partitioning heuristic always selects the exact same set of activities for elimination, since it now relies on the *ei* calculations. Since calculations of residuals are invariable, the created partitions would be exactly the same for every simulation run.

Secondly, the stopping strategy has been proposed in the way as described earlier to solve two other limitations:


Finally, the design of two different options (standard or advanced) for the selection and the stopping strategies is new and solves the last and most important limitation, as follows:

• **Solution 3.** The extended version of the calibration procedure added project data partitioning as a promising feature to accept lognormality, but this new feature could only be performed based on managerial criteria influenced by human judgement. The statistical partitioning heuristic has followed the same logic, but transformed it into a statistical, rather than managerial, partitioning approach. Statistical partitioning is not subject to human (mis-)judgement and not victim to human biases but does not exclude the option of human partitioning as an initialisation step (S0). In the computational experiments of Section 5, it will be shown that human and statistical partitioning lead to a higher acceptance rate of project data.
