Object Reference : Object View and Procedure Reference : Series
  
 
outliers
Detect outlying observations.
Use Tukey fences, mean/standard deviation fences, wavelet outliers, or ARMA outliers detection methods to identify observations that deviate from the natural pattern of the underlying series.
Syntax
series_name.outliers(options)
Options
 
hp
Perform the Hodrick-Prescott filter on the series, and then detect outliers on the cycle series.
diff
First-difference the series before detecting outliers.
sens=arg
Set the sensitivity level. Valid arguments are “low”, “medium” (default), and “high”.
nofence
Do not perform Tukey and mean/standard deviation fences.
nowave
Do not perform Wavelet Outlier detection.
arma
Perform ARMA based outlier detection (this option is turned on by default in dated workfiles).
noarma
Do not perform ARMA based outlier detection (this option is turned on by default in undated workfiles).
tukeyk=arg
Set the value k in the Tukey fence detection routine. This will override the value of k set by the sens= option.
meanstdevk=arg
Set the value k in the mean/standard deviation fence detection routine. This will override the value of k set by the sens= option.
wavesig=arg
Set the value false discovery rate significance value used in the Wavelet Outlier detection routine. This will override the value set by the sens= option.
armac=arg
Set the value c in the ARMA outlier detection routine. This will override the value of c set by the sens= option.
setsmpl
Set the workfile sample to be any observations identified as outliers.
excludesmpl
Set the workfile sample to be any observations identified as not outliers.
series=name
Create a new series in the workfile, named name, containing a value of 1 for any observations identified as an outlier, and a value of 0 for any observation identified as not an outlier.
datestring=name
Create a new string object in the workfile containing the dates (or observation identifiers) for any observations identified as an outlier.
nogrlabels
Turn off observation labels on the outlier graph.
prompt
Force the dialog to appear from within a program.
p
Print results.
Examples
gdpc1.outliers(hp, nowave, sens=high, meanstdevk=3)
Performs outlier detection on the cycle series from a Hodrick-Prescott filter of the GDPC1 series, opting to not use Wavelet Outlier detection, and setting the sensitivity of the detection to "high", but overwriting the k value used in the mean/standard deviation fence to 3.
Cross-references
See “Outlier Detection”. See also Equation::fitoutliers.