IMPORTANT! Goodness-of-fit metrics

[cvallee]

Fit metrics to evaluate the goodness-of-fit of the fit.

Current ideas:

• sqrt(Chi^2): same metrics as Creoptix
• Separated Chi2 values for association and dissociation pulse (residual weights on association phases is set to 0, so the fit at the association phases should be worth than the fit at the dissociation phases)
• RMSE at the start of the dissociation (sharp decrease) vs RMSE at the elbow

[cvallee]

Sqrt(Chi2) implemented in #20

[cvallee]

Connor says Chi2 isn't a great metrics for GoF because it assumes a defined distribution of the residuals, which might not be respected in our case. He suggested to use lilliefors test approach instead which is less affected by prior assumption