Full text loading...
-
A Comparison Between Robust M-Estimation And Adaptive Lp-Norm Techniques In Determining The Best Apparent Resistivity Versus Frequency Curves In Mt Soundings
- Publisher: European Association of Geoscientists & Engineers
- Source: Conference Proceedings, 5th SAGA Biennial Conference and Exhibition, Sep 1997, cp-223-00027
Abstract
Data collected at a magnetotelluric sounding station are usually severely affected<br>by noise. The result is that impedance estimates do not fall on a single point but<br>are scattered. Statistical manipulation of the data is necessary to reduce the<br>scattered points to the most probable value. Although least squares regression<br>provides a simple way of doing this it can lead to erroneous results since outliers<br>ensure that the residuals are not normally distributed. Robust M-estimation deals<br>with outliers by applying a weight function, determined by normalized values of<br>the residuals, to the least squares regression formula. Contrary to this the adaptive<br>Lp norm technique uses the quality of the data to determine the best norm value<br>for minimizing the error function. The results of tests with synthetic data show<br>that the adaptive Lp norm technique yields the best results.