Fig. 13
A comparison of the spread of input parameter locations of the acceptable runs found for the k 2c >0 model (red boxplots) and the k 2c =0 model (blue boxplots) in terms of individual inputs as labelled along the x-axis. Note that the two sets of acceptable runs being compared correspond to the red lines in Figs. 10 and 12 respectively. The y-axis is on a log10 scale, and the grey rectangles show the initial ranges that define the original search region \(\mathcal {X}_{1}\) as given in Table 2. The light blue horizontal lines show the input parameter values of the previous best run as found by [30]. We can see that some inputs, for example, k 1, are either unconstrained by the observed data, or possibly that changes in these inputs can be compensated by appropriate changes in other inputs. Some inputs such as k 16a /k 16 are highly constrained, while others such as k 2/k 1a , k 2b , k 3/k 1a and k 3a /k 1a show clear differences between the k 2c >0 and k 2c =0 models, with the latter model preferring higher values of k 2/k 1a and k 3a /k 1a and lower values of k 2b and k 3/k 1a