Object Reference : Object View and Procedure Reference : Series
  
 
autoarma
Forecast from a series using an ARIMA model with the specification of the model selected automatically.
Syntax
series.autoarma(options) forecast_name [exogenous_regressors]
Options
 
tform=arg
Specify the type of dependent variable transformation. arg may be “auto” (automatically decide between log or no transformation, default), “none” (perform no transformation), “log” (perform a log transformation), and “bc” (perform the Box-Cox transformation.
bc=int
Set the power of the Box-Cox transformation. Only applicable if the tform=bc option is used.
diff=int
Set the maximum level of differencing to test for. Default value is 2.
maxar=int
Set the maximum number of AR terms. Default value is 4.
maxma=int
Set the maximum number of MA terms. Default value is 4.
maxsar=int
Set the maximum number of seasonal AR terms. Default value is 0.
maxsma=int
Set the maximum number of seasonal MA terms. Default value is 0.
periods=int
Set the periodicity of the seasonal ARMA terms. This defaults to the number of observations in a year, based on current workfile frequency.
avg=key
Use forecast averaging, rather than model selection. key sets the type of averaging to perform, and may take values of “aic” (SAIC weights), “sic” (BMA weights) or “uni” (uniform weights).
select=key
Set the model selection criteria. key make take values of “aic” (Akaike Information Criteria, default), “sic” (Schwarz Information Criteria), “hq” (Hannan-Quinn criteria) or “mse” (Mean Square Error criteria). This option is ignored if the “avg=” option is used.
nonconv
Allow non-converged models to be used in model selection or forecast averaging.
mselen=key
Set the percentage of the estimation sample to be used for MSE calculation. key may take values of “5”, “10”, “15” or “20”. This option is only applicable if the “select=mse” option is used.
msetype=key
Set the type of forecast to use when calculating MSE. key may either be “dyn” (dynamic, default), or an integer, n, between 1 and 12 indicating that an n-step static forecast should be performed. This option is only applicable if the “select=mse” option is used.
kpsssig=key
Set the significance level of the KPSS test when determining the appropriate level of differencing for the dependent variable. key may take values of “1”, “5” (default) or “10”.
fgraph
Output a forecast comparison graph.
atable
Output a selection criteria comparison table
agraph
Output a selection criteria comparison graph.
etable
Output a final equation output table. Not applicable if the “avg= option is used.
eqname=name
Create an equation object in the workfile with the same specification as the final selected equation. Not applicable if the “avg=” option is used.
seed=num
Set the random number generator seed for random starting values.
prompt
Force the dialog to appear from within a program.
p
Print results.
Forecast sample options
The forecast sample will start at the observation immediately after the estimation sample (the current workfile sample). The forecast endpoint is given by either:
 
forclen=int
Number of periods to forecast.
forc="date"
Specify the date of the forecast end point.
If omitted, the end point will be the end of the workfile range.
Example
The commands
wfopen elecdmd.wf1
elecdmd.autoarma(maxsar=1, maxsma=1, noconv, forclen=20, agraph, atable, fgraph) elecdmd_f @expand(@month) realgdp tempf
open the workfile “elecdmd.WF1” and then perform automatic forecasting on the series ELECDMD. The forecasts will be stored in a series called ELECDMD_F. The ARIMAX model includes exogenous regressors of REALGDP, TEMPF and a set of monthly dummy variables, created with the @expand keyword.
The number of maximum SAR terms and SMA terms are set to 1 (instead of the default 0). Model selection is used to determine the best ARMA model, with non-converged models included in the selection process.
The forecast covers 20 periods, and upon completion, EViews will display a graph of the Akaike information criteria of each of the ARMA models considered, as well as a table of each of the selection criteria, and a graph of the each of the forecasts.
Cross-references
See “Automatic ARIMA Forecasting” for additional discussion.