Least Absolute Regression Revisited |
| |
Authors: | Paul Rech Paul Schmidbauer Jamie Eng |
| |
Institution: | School of Business , San Francisco State University , 94132, San Francisco, CA |
| |
Abstract: | We describe a method for fitting a least absolute residual (LAR) line through a set of two–dimensional points. The algorithm is based on a labeling technique derived from linear programming. It is suited for interactive data analysis and can be carried out with graph paper and a programmable hand calculator. Tests conducted with a Pascal program indicate that the algorithm is computationally efficient. |
| |
Keywords: | least absolute regression algorithm L1 estimation resistant line |
|