Multiple linear regression curve fitting: A quadratic programming solution. (English) Zbl 0841.65150
Florenzano, Monique (ed.) et al., Approximation and optimization in the Caribbean II. Proceedings of the 2nd international conference held in Havana, Cuba, September 26 - October 1, 1993. Frankfurt a. M.: Peter Lang. Approximation Optimization. 8, 1-13 (1995).
Summary: The problem of fitting a regression curve is solved by using a convex combination of least squares and least absolute deviation. This method is robust against the existence of outliers. A one-parameter quadratic programming problem is derived for solving the fit of the model. A classical problem is analyzed using the proposed algorithm.
These results sustain that the proposal has an adequate performance in terms of the accuracy of the estimates.
For the entire collection see [Zbl 0836.00031].
These results sustain that the proposal has an adequate performance in terms of the accuracy of the estimates.
For the entire collection see [Zbl 0836.00031].
MSC:
| 65C99 | Probabilistic methods, stochastic differential equations |
| 65D10 | Numerical smoothing, curve fitting |
| 90C20 | Quadratic programming |
| 62J05 | Linear regression; mixed models |
