outliers - Detect and optionally fix outliers
SYNOPSIS
outliers [ parameter=value ... ] [ inputfile ... ]
Parameters are: include_vars, box_dims,
min_good_vals, threshold, fix_outliers.
DESCRIPTION
outliers detects outliers using the following rule:
abs(x - median(x)) > threshold
where the median is computed in a box of user-specified size around x.
A user-specified minimum number of good elements must exist in a box before the median
can be computed for the box. If the number of good points does not exceed this minimum
value, the value at the center of the box is assumed not to be an outlier.
All variables to be processed must have the same number of dimensions, either one or
two.
Outliers can either be replaced by the medians for the corresponding boxes, or by the
missing value code.
PARAMETERS
- include_vars
- List of variables to process. If the list is preceded by a minus sign, then all
variables except those listed will be processed. Wildcards * and ?
are allowed. The default is to process all variables.
- box_dims
- One or two numbers specifying the size of the box. The box must have the same number of
dimensions as the variables to be processed. The valid range is [1,50]. There is no
default.
- min_good_vals
- The minimum number of good values needed in the moving box to allow the median to be
computed. The valid range is 1 up to the number of elements inside the box. The
default is 1 if the box is 1-D, or the height of the box if the box is 2-D.
- threshold
- The maximum acceptable value in the outlier detection rule. The valid range is [ >=
0 ]. There is no default.
- fix_outliers
- If yes, specifies that an outlier is to be replaced by the median computed for
the corresponding box. If no, outliers are replaced by the missing (bad) value
code. The default is yes.
SEE ALSO
smear, nhood
Last Update: $Date: 1998/05/29 20:13:07 $