We also studied the efficiency of the reconstruction method:
When the number of iterations was restricted to 50z (see Section
2.1), the range of final energy values was between 102.8
and 1010.1. 525 plot reconstructions (74%) had energy values
smaller than 106.
The alternative included setting an energy limit of 106 and restarting
the reconstruction if the number of iterations were in access
of 10,000 (see Section 2.1). It turned out that only in 39 cases
(5.5%) the reconstruction simulations had to be re-set and started
again. The range of final energy values of the 5.5% re-started reconstruction
simulations was between 104.6 and 105.9, i.e. some of
those were close to acceptance in the first simulation.
Interestingly the correlations between observed and reconstructed
structural summary characteristics vary considerably.
The mean directional index has a correlation coefficient of 0.23,
the diameter differentiation one of 0.64, and the correlation coefficients
of the distance to the nearest neighbour and mingling are
0.66 and 0.83, respectively. These values disagree with the weights
of the energy function specified in Section 2.1.1. With a F value of
1.17 the highest ratio of variance exists between observed and
reconstructed nearest neighbour distance.