Robust regression and its application in absolute gravimeters.

Maximum-likelihood estimation of robust regression is employed to process absolute gravimeter data because this method is insensitive to outliers. Gravity values obtained by both ordinary least squares and different approaches for robust regression are compared with the results obtained by the g9 processing software for the FG5 gravimeter. The results show that the accuracy of gravity data can be improved by rejecting certain outliers via adjustments to a suitable weighting function of time-distance pairs. A convenient method of identifying those outliers by their weight is also presented.