This panel helps to quantify the global influence of model parameters on the observed outcomes. The Simmune Analyzer uses a global variance based method to measure the relative influence of a parameter over the whole support of its distribution. This is done by doing an ANOVA decomposition of the observed variance of a simulation output into contributions due to different parameters. In order to do the decomposition a model has to have at least two parameters that can be varied. The output will be so called Sobol Indices, which measure the relative contribution of a parameter to the variance. The sum of all contributions is smaller than, or equal to 1 (except for errors due to a finite sample size). If the sum is smaller this is an indication of interactions between two or more parameters. Parameters that have a large Sobol Index at experimentally relevant times are good candidates for measuring their value since reducing them has a large probability to reduce the model prediction's variance. The Simmune Analyzer also offers to view the total Sobol indices of an observable, these capture not only the effect of a single parameter, but also all of it's interactions, consequently the sum of the total Sobol Indices is larger than 1, except for the case of no interactions, in which it coincides with the Sobol Index. Total Sobol Indices are useful to identify unimportant parameters. If the total Sobol Index of a parameter is small its value can be fixed anywhere within its range without much affecting the model response. If the coefficient of variation of a variable is small the estimation the its Sobol indices becomes sensitive to round-of errors. Therefore, no Sobol indices (total or otherwise) will be displayed if the coefficient of variation is smaller than 1e-4.

The estimation of Sobol indices requires a specialized sampling strategy. Select a previously created setup as a template to create a new setup for the evaluation of the Sobol Indices. The actual number of simulation that will be run will be larger than the number of samples chosen, additionally to the normal evaluation one extra point plus two additional points per parameter have to evaluated. The number displayed is the effective number of samples for the statistical estimation of the Sobol Indices. By default the sobol indices of observables at the start of a simulation are shown, when you enter a time value to display the Sobol Indices at different times, the time you entered will be rounded to the nearest time stored in the database. Alternatively you can choose to plot the parameter sensitivities of a single observable over time by checking the Plot sensitivity time series box, choose the observable of interest by clicking in the row or column in the table above the plot, which is displaying the numerical values of the sensitivities.