Format of simulation tables #280
Description
I have the feeling the current simulation tables are less helpful than they could be. My understanding is that they are there to 1) facilitate assessing the goodness of fit for the provided nominal parameters without the need for simulation, and 2) be used as test oracle.
Currently, the first objective is complicated by the noiseParameters just being copied from the measurement tables. In the best case, this requires substitution of parameters from the model, parameter table, or condition table; in the worst case, this requires re-simulation, because the error model is state-dependent. Therefore, I think it would be preferable to include the nominal values of the distribution parameters. See also the discussion at PEtab-dev/PEtab#632.
Another pain point is matching the measurements and simulation table rows (we can't rely on the same ordering). If all information from the measurement table is included in the simulation table, the rows can be matched. However, this could be greatly simplified by including some measurement ID in both the measurement and simulation table.
Partially addressing the first issue, some simulation tables have numeric observableParameters and noiseParameters instead of the original values from the measurement table. However, this further complicates matching simulations and measurements.