OUTLIER Subcommand (TSMODEL command)

The OUTLIER subcommand specifies outlier time points to be included in an ARIMA model.

  • By default, no time points are modeled as outliers.
  • For each outlier, you must specify its time point (location) and type.
  • Multiple OUTLIER subcommands can be used to define two or more outliers for an ARIMA model. Duplicate specifications (same location and type) are ignored and a warning is issued.
  • A warning occurs and the OUTLIER subcommand is ignored if the subcommand is part of a MODEL block that contains an EXPERTMODELER or EXSMOOTH subcommand.
  • If the AUTOOUTLIER and OUTLIER subcommands are both included in a MODEL block containing an ARIMA subcommand, a warning is issued and the OUTLIER specification takes precedence.

Example

TSMODEL
  /MODELDETAILS PRINT=PARAMETERS
  /MODEL DEPENDENT=revenue
  /ARIMA AR=[1] MA=[1]
  /OUTLIER LOCATION=[YEAR 2000 MONTH 8] TYPE=LOCALTREND.
  • A custom ARIMA(1,0,1) model is specified with an outlier of type Local Trend for August, 2000.

LOCATION Keyword

The LOCATION keyword specifies the time point to be treated as an outlier and is required.

  • Specify a date in square brackets. If the data are undated, you must specify a case number. Otherwise, specify the outlier location using date keywords and values. For example, LOCATION=[YEAR 2000 MONTH 8] specifies that August, 2000 be treated as an outlier. The following date keywords may be used: CYCLE, YEAR, QUARTER, MONTH, WEEK, DAY, HOUR, MINUTE, SECOND, and OBS. Every keyword in the IBM® SPSS® Statistics date specification must appear in LOCATION. Only keywords that correspond to system date variables may be used, and no date keyword may be used more than once. See the topic DATE for more information.
  • A warning is issued if you specify an outlier location that is outside the estimation period or corresponds to a gap in the data. Any such models are ignored.

TYPE Keyword

The TYPE keyword specifies the type of outlier and is required. Specify one of the following types:

ADDITIVE. Additive. An outlier that affects a single observation. For example, a data coding error might be identified as an additive outlier.

LEVELSHIFT. Level shift. An outlier that shifts all observations by a constant, starting at a particular series point. A level shift could result from a change in policy.

INNOVATIONAL. Innovational. An outlier that acts as an addition to the noise term at a particular series point. For stationary series, an innovational outlier affects several observations. For nonstationary series, it may affect every observation starting at a particular series point.

TRANSIENT. Transient. An outlier whose impact decays exponentially to 0.

SEASONALADDITIVE. Seasonal additive. An outlier that affects a particular observation and all subsequent observations separated from it by one or more seasonal periods. All such observations are affected equally. A seasonal additive outlier might occur if, beginning in a certain year, sales are higher every January. The keyword generates an error if the season length has not been defined via the DATE command or the SEASONLENGTH keyword.

LOCALTREND. Local trend. An outlier that starts a local trend at a particular series point.