1.0 - 8.00 - VARMAX Arguments - Teradata Vantage

Teradata® Vantage Machine Learning Engine Analytic Function Reference

Product
Teradata Vantage
Release Number
1.0
8.00
Release Date
May 2019
Content Type
Programming Reference
Publication ID
B700-4003-098K
Language
English (United States)
ResponseColumns
Specify the columns containing the response data. Null values are allowed at the end of the series. If you specify StepAhead, the function reports predicted values for the missing values, using values from the predictor columns for those time periods.
ExogenousColumns
[Optional] Specify the columns containing the exogenous (independent) predictors.
Default behavior: The function calculates the model without exogenous vectors.
PartitionColumns
[Optional] Specify the partition columns to pass to the output.
PDQ
[Required if orders_table is omitted, disallowed otherwise.] Specify the parameters p, d, and q for the VARMA part of the model. These values must be nonnegative integers in the range [0, 20].
SeasonalPDQ
[Optional if orders_table is omitted, disallowed otherwise.] Specify seasonal parameters sp, sd, and sq for the VARMA part of the model. These values must be nonnegative integers in the range [0, 20].
If you specify this argument, you must also specify the Period argument.
Default behavior: The function treats the model as nonseasonal.
Period
[Optional if orders_table is omitted, disallowed otherwise.] Specify the period of each season. The period must be a positive INTEGER value.
If you specify this argument, you must also specify the SeasonalPDQ argument.
Default behavior: The function treats the model as nonseasonal.
ExogenousOrder
[Optional] Specify the order of the exogenous variables. If the current time is t and ExogenousOrder is b, the following values of the exogenous time series are used in calculating the response: X t X t-1 ... X t-b+1.
Default behavior: The function calculates the model without exogenous vectors.
Lag
[Optional] Specify the lag in the effect of the exogenous variables on the response variables. For example, if lag is 3, and ExogenousOrder is b, the prediction Y i is based on X i-3 to X i-b-2.
Default: 0
IncludeMean
[Optional if orders_table is omitted, disallowed otherwise.] Specify whether to add the mean vector of the response series in the VARMAX model.
If this argument is 'true', the values of the difference parameters d (in the PDQ argument) and sd (in the SeasonalPDQ argument) must be 0.
Default: 'false'
IncludeDrift
[Optional if orders_table is omitted, disallowed otherwise.] Specify whether to add the drift term of the response series in the VARMAX model. The drift term applies only when d is 1.
IncludeDrift and IncludeMean cannot both be true. If either the arguments or order table specifies true for both IncludeDrift and IncludeMean, the function issues an error message.
Default: 'false'
MaxIterNum
[Optional] Specify the maximum number of iterations performed. The max_iteration_number must be a positive INTEGER value.
Default: 100
StepAhead
[Optional] Specify the number of steps to forecast after the end of the time series. The predict_steps must be a positive INTEGER value.
Default behavior: The function calculates no forecast values.
Method
[Optional] Specify the method for fitting the model parameters:
Option Description
'SSE' (Default) Sum of squared error.
'ML' Maximum likelihood. This method is more accurate, but slower to calculate.