# Classify Statistical Outliers

**Classify Statistical Outlier** is another type of noise classifier. This routine first creates a surface from a neighborhood of points (**Neighbor Count**). Then, residuals in respect to this surface are computed for each point. Any points with residuals that fall outside of the sigma specified (**Sigma Multiplier**) are assigned to the **Output Class**.

In the example below, any points with a residual larger than sigma 1.0 are assigned to the Low Point (noise) class:

<figure><img src="https://2222094320-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2FdEevfLZRIk38LUPwDa4V%2Fuploads%2Fypvbl2Wl5g4PrYUnjsqt%2Fimage.png?alt=media&#x26;token=b5ae3446-19ff-4485-86ff-88fe96c722d5" alt=""><figcaption></figcaption></figure>

This tool is ideal for cleaning high density data sets and for tightening up planar surfaces. In general, the **Sigma Multiplier** parameter controls how aggressively this tool classifies points as noise. Below are some example values:

| Sigma Multiplier Value | Use-case                                                                                                                         |
| ---------------------- | -------------------------------------------------------------------------------------------------------------------------------- |
| 3.00                   | Will classify very few points as noise. Ideal for high-precision point clouds.                                                   |
| 1.00                   | Will classify a moderate amount of points as noise, and generally a good starting point when experimenting with outlier removal. |
| 0.5                    | Will classify about half of the data set as noise, and useful when attempting to drastically improve low precision data sets.    |