1. Home
2. TUKEYBIWEIGHT procedure

# TUKEYBIWEIGHT procedure

Estimates means using the Tukey biweight algorithm (D.B. Baird).

### Options

`CUTPOINT` = scalar Cut point after which weight is set to zero; default 5 Tolerance to avoid division by zero; default 0.00001

### Parameters

`DATA` = variates or pointers Data values Groupings of the data values Saves the means Saves standard errors

### Description

`TUKEYBIWEIGHT` estimates means using the Tukey biweight algorithm. This weights the data values depending on how far they are from the median, and discards any that are more than a specified number of median absolute distances away. The number of differences is specified by the `CUTPOINT`, with a default of 5.

The data values are specified by the `DATA` parameter. They can be in a single variate, with any groupings specified by the `GROUPS` parameter. Alternatively, they can be in separate variates, one for each group. The `MEANS` parameter saves the estimated means, and the `SE` parameter saves standard errors for the means.

Options: `CUTPOINT`, `TOLERANCE`.

Parameters: `DATA`, `GROUPS`, `MEANS`, `SE`.

### Action with `RESTRICT`

`TUKEYBIWEIGHT` respects any restrictions on `DATA` or `GROUPS`.

Procedures: `MPOLISH`, `ROBSSPM`.

Commands for: Calculations and manipulation.

### Example

```CAPTION       'TUKEYBIWEIGHT example','Mean vs. Tukey BiWeight';\
STYLE=meta,minor
FACTOR        [LEVELS=100; VALUES=(1...100)100] Group
"Create X as a mixture of 2 normals, 10% with SE of 10 vs 1"
CALCULATE     [SEED=498725] X1,X2 = GRNORMAL(9000,1000;10;1,100)
APPEND        [NEWVECTOR=X] X1,X2
TABULATE      [CLASS=Group] X; MEANS=AMeans
VTABLE        AMeans; VARIATE=Average
TUKEYBIWEIGHT [CUTPOINT=3] X; GROUPS=Group; MEANS=TukeyBiWt
DHISTOGRAM    [TITLE='Distribution of Mean vs Tukey Biweight'; NGROUPS=15]\
Average,TukeyBiWt
```
Updated on March 4, 2019