1: \begin{abstract}

2:   The robust adjustment of nonlinear models to data is considered in

3:   this paper. When data comes from real experiments, it is possible

4:   that measurement errors cause the appearance of discrepant values,

5:   which should be ignored when adjusting models to them. This work

6:   presents a Lower Order-value Optimization (LOVO) version of the

7:   Levenberg-Marquardt algorithm, which is well suited to deal with

8:   outliers in fitting problems. A general algorithm is presented and

9:   convergence to stationary points is demonstrated. Numerical results

10:   show that the algorithm is successfully able to detect and ignore

11:   outliers without too many specific parameters. Parallel and

12:   distributed executions of the algorithm are also possible, allowing

13:   for the use of larger datasets. Comparison against publicly

14:   available robust algorithms shows that the present approach is able

15:   to find better adjustments in well known statistical models.

16: \end{abstract}

17: