Heteroscedasticity-consistent covariance matrices (HCCM) have been implemented in the new private\hccm.m function and used by the snnls.m solver. There, the HC1 estimation is used, owed to its popularity and extended usage in other fields.
HCCM allows for the proper estimation of the covariance matrices when doing global fits, i.e. when the augmented LSQ residual vector contains errors of different variance. Since the HCCM estimates are compatible with homoscedasticit residuals, this would lead to a general and simplified code.
Therefore, it should be implemented in all other fit functions which use global fitting and compute covariance matrices from LSQ residuals fitparamodel,fitsignal,fitregmodel,...
luisfabib
commented
4 years ago
Heteroscedasticity-consistent covariance matrices (HCCM) have been implemented in the new
private\hccm.mfunction and used by thesnnls.msolver. There, the HC1 estimation is used, owed to its popularity and extended usage in other fields.HCCM allows for the proper estimation of the covariance matrices when doing global fits, i.e. when the augmented LSQ residual vector contains errors of different variance. Since the HCCM estimates are compatible with homoscedasticit residuals, this would lead to a general and simplified code.
Therefore, it should be implemented in all other fit functions which use global fitting and compute covariance matrices from LSQ residuals
fitparamodel,fitsignal,fitregmodel,...