Fitting weights

[damskii9992]

Background

When fitting data to a model it is possible to add weights to different data-points to prioritize the fit to certain parts of the data which might be more important or more accurate than other parts. This is especially useful when working with data with uncertainties, as it is possible to weigh more accurate data-points higher than more inaccurate points.

The Decision

Using any default weights, such as 1, i.e. weighting all data evenly, can cause a subtle mistake in the estimation of the errors of the fitted parameters. It is therefore decided to not have any default weights at all and require that users supply a list of weights in all cases.
It is recommended that this list be the variances of the data, if available.

This decision is enforced in the core-library but can be overwritten by technique-specific libraries.

We also decided to have an internal check to check that no weights are 0 or Inf / NaN to avoid errors.

Code details

Certain minimizers expect to be supplied with 1/weights, others simply with weights, we have handled this internally and users are expected to simply supply the variances.

Link to the ADR suggestion:
#135