Automatic best‐guess and Error Estimation

One of the most difficult aspects of the scattering deconvolution process is getting a reasonable estimate of the uncertainties.

Currently implemented

Using the smoothed third derivative of the figures of merit computed for each tau value trial, the first peak in that timeseries nominally corresponds to the value at which over-subtraction begins (assuming that the range of taus searched do actually include the real value).

An error can be constructed from the quadrature sum of the tau trial spacing and the standard deviation of the set of metric best-guesses. It is not exactly "robust" as there's not theoretical underpinning for why that quantity necessarily encodes the uncertainty associated with the deconvolution. In fact, this is a long standing problem with the CLEAN method in synthesis imaging, too... an "error" is not well defined.