|
Abstract:
|
All the existing regression estimates suffer from scale problem that exaggerate the contribution of extreme observation(s) as well as outlier(s). Unlike the traditional fitting procedures of regression line based on the least squares estimates; it uses one dimensional transformed deviation for minimizing the total sum of errors. Interestingly enough it is notified that, the estimators have been made a link to the ratio estimators. These estimators also suggest a new approach of quantifying the quality of estimation in the presence of abridged errors, detecting outlier, etc in regression analysis.
|
