Hi Olaf
So I think the easiest way to achieve what you want is to use multiple forward models and put all measurements in one single data group. You will then have to modify the .MOMap (Section 2.5.3 of the Bayesian inversion manual) to properly match the individual data points to the correct models and model outputs. To get around the issue of non-uniform number of measurements, just arrange all measurements in a row vector.
As an example, assume you have two models with scalar outputs and three data points where the first data point is assigned to the first model’s output and the second and third to the second model’s output. You should then create a single data group like:
myData.y = [y_1, y_2, y_3]; myData.MOMap = [1 2 2;... 1 1 1];
You will then also need to provide a discrepancy model to go with this data group.
Please let me know if this is clear and if it solves your problem. To double check that everything is set up correctly, you can start the debugger inside the uq_inversion_likelihood function. This is where the data and model outputs are compared.