The results of the PPSM simulation give the possibility to identify the supplementary costs generated by the architecture optimization meant to eliminate the flow concentrators.
The new results are transferred as data entry for PPSM activity flow in order to see if the financial gain generated by the increased productivity covers the investment value.
The effects of the available explanatory variables, PCI and PPSM, are modeled semiparametrically.
Following Frees and Wang (2005), a rescaled version of PCI is used, namely, PCI/1000, together with the logarithm of PPSM.
Second, a lognormal mixed model is fitted to the data, in which nonlinear effects of both PCI and PPSM are allowed.
Modeling nonlinear effects of PCI and PPSM, the combination of random intercepts and a nondiagonal [[summation].