Multi-model inference and model uncertainty quantification

[luisfabib]

The current design of the fit function would allow for a "simple" implementation of multi-model inference and of the related model-uncertainty (both in a statistical-theoretical as a bootstrapped framework).

In terms of interface, it should be relatively easy. One defines multiple models that might describe the data model1 to modelN and these are passed as a list to the fit function

models = [model1,...,modelN]
results = dl.fit(models,data)

The uncertainty propagation through bootstrap is already almost implemented in DeerLab and the theoretical approximate approach would require just a small additional calculation to the current uncertainty engine.

Furthermore, one could imagine having options such as multimodel='selection' to perform model selection or multimodel='average' to work with the model-averaged fit during the multi-model inference.

All the math and concepts to easily implement this can already be found in the Burnham & Anderson 2002 book.