**Syntax**

` >>> ANDREW: c`

**Parent Command**

` >> OPTION`

**Subcommand**

` -
`

**Description**

This command selects a robust estimator named Andrews. Given this estimator, the objective function to be minimized is the sum of the cosine functions g(y), where y is the weighted residual:

where:

with:

This objective function does not correspond to a standard probability density function. It has the general characteristic that the weight given individual residuals first increases with deviation, then decreases to reduce the impact of outliers. The parameter *c* indicates the deviation at which residuals are considered to be outliers. If the measurement errors happen to be close to a normal distribution with standard deviation sigma, then the optimal value for the constant *c* is *c*=2.1. Note that this objective function can be minimized using the standard Levenberg-Marquardt algorithm which is designed for a quadratic objective function. Since (1-cosine) can be reasonably well approximated by a quadratic function for small *y _{i}*, the Levenberg-Marquardt algorithm is usually quite efficient.

**Example**

` > COMPUTATION
>> OPTION
>>> use the robust estimator ANDREW with a constant `

*c*: 1.5 <<< <<

**See Also**

` >>> CAUCHY | >>> L1-ESTIMATOR | >>> LEAST-SQUARES | >>> QUADRATIC-LINEAR
`