X12arima for OxMetrics 4

Contents:

Overview
Introduction
X-12-ARIMA
Disclaimer
Limitations
Documentation
Getting started
Starting OxMetrics and X12arima
X12arima Commands
Quick Seasonal Adjustment
Formulate, Estimate and Diagnostic Graphics
ARIMA model example
ARIMA model example
regARIMA model example
Batch usage
Batch Usage
Additional Batch Commands
Specification Syntax, Additions and Differences

Overview

Introduction

The X12arima for OxMetrics 4 program gives access to most of the X-12-ARIMA features for seasonal adjustment and regARIMA modelling. Variables are selected from a OxMetrics database, and an X12 specification is formulated using dialogs. After successful estimation, the results can be graphed, or stored in a OxMetrics database. OxMetrics graphs can be manipulated on screen, and saved to variety of formats.

X-12-ARIMA

X-12-ARIMA is the most recent outcome of a research programme on seasonal adjustment which has been undertaken by researchers at the US Census Bureau since the 1950's. At the heart is the X-11 seasonal adjustment method, which is widely used by statistical agencies throughout the world.

X-12-ARIMA builds on X-11-ARIMA, adding the ability to extend the time series with forecasts and backcasts from ARIMA models prior to seasonal adjustment.

The current version, called X12arima for OxMetrics, uses the code created and made available by the Census Bureau, with minor modifications. The Census version works through a specification file (.spc), which sets the options and determines which actions to take. The OxMetrics version preserves the specification syntax, and builds a specification through a set of dialogs. Existing .spc files can be used as OxMetrics batch files for X12arima for OxMetrics after minor modifications. It is not necessary to first create .spc files: seasonal adjustment models can be formulated entirely interactively, with the data loaded in OxMetrics.

The X-12-ARIMA method modifies the X-11 variant of Census Method II by J. Shiskin A.H. Young and J.C. Musgrave of February, 1967, and the X-11-ARIMA program based on the methodological research developed by Estela Bee Dagum, Chief of the Seasonal Adjustment and Time Series Staff of Statistics, Canada, September, 1979. Primary Programmers: Brian Monsell, Mark Otto.

We suggest that you cite

Findley, D.F., Monsell, B.C., Bell, W.R., Otto, M.C. and Chen, B-C (1998), `New Capabilities and Methods of the X-12-ARIMA Seasonal Adjustment Program'. Journal of Business and Economic Statistics, Vol 16.
whenever reporting results from X12arima for OxMetrics.

Disclaimer

No warranty whatsoever is given for these programs.
You use them at your own risk!

Limitations

Documentation

Full X-12-ARIMA documentation is downloadable. Specifically:

Getting Started

Starting OxMetrics and X12arima

Start OxMetrics from the taskbar (or from the OxMetrics group or icon on your desktop). To start X12arima, access the Run menu in OxMetrics, or start X12arima from the workspace window. If X12arima is not listed, you must exit OxMetrics, and install X12arima.
X12arima can only be started from OxMetrics!

The X12arima program will look like this:

First load the ICMETI.in7 tutorial data set, which holds the monthly series entitled `Total Inventories Communications Equipment' from 1968M1 to 1989M3. If you installed in the default directory structure, the data will be in the \Program Files\OxMetrics4\X12arima directory


X12arima Commands

The X12arima program can be used in two modes. The first is entitled `Quick seasonal adjustment', where all settings are kept at their defaults. The second mode allows for detailed specification. Before we give an example of both, here is a summary of the menu entries:

Model Menu Commands

Test Menu Commands


Quick Seasonal Adjustment

Select Quick Seasonal Adjustment from the X12arima model menu, or click on the Q icon on the toolbar. Make sure the ICMETI database is selected, and double click on the ICMETI variable to add it to the list of selected variables (or click once to select it, and then use the Add button):

Press OK to run the default X11 seasonal adjustment procedure. The graphic analysis dialog pops up automatically after quick seasonal adjustment, select all graphs:

The output from each run appears in a text window inside OxMetrics, entitled `X12arima [Console]' (the [Console] part is removed when closing the X12arima module). In Quick adjustment mode, only summary output is displayed. This is much more compact than the default from X-12-ARIMA. When using the normal estimation mode, it is possible to select other output mode (some of which will generate thousands of lines of output for each series).

The Diagnostics Graphics entry on the Test menu will be considered in the next section.


Formulate, Estimate and Diagnostic Graphics

Much more control is available when using Formulate and Estimate from the Model menu. As an example, we select the log transformation for the ICMETI variable. Select Formulate; ICMETI is still selected (if not, use the button labelled Select):

Press Transform and select the log transform:

Press OK, and then the Estimate button (or, equivalently, close the X12-ARIMA dialog and use Estimate from the Model menu). After estimation, the entries on the Test menu become active, as well as their corresponding toolbar buttons. (Unlike for quick estimation, the graphic analysis dialog does not appear automatically.) Diagnostic graphics can be used to graph the spectra of the (differenced) seasonally adjusted variable, the (differenced) original, and the irregular:

ARIMA model example

Introduction

The X12arima for OxMetrics program can be used for ARIMA modelling, either with or without using the seasonal adjustment features. As an example, we first estimate the airline model, followed by a regARIMA model (ARIMA model with regressors).

Formulating the ARIMA model

Load the airline.xls data set in OxMetrics, and start X12arima (if it is not already active). Select Model/Formulate, or click on the model formulation icon on the toolbar.

The airline model is:

(1-L)(1-L12)log(yt)= (1-b1L)(1-b12L12)et.

The following steps are required:

The resulting main dialog will look like this:

Model Output

The model output is available graphically:

Numerical results are reported in OxMetrics:

MODEL DEFINITION for Airline Transformation: Log(y) ARIMA Model: (0 1 1)(0 1 1) regARIMA Model Span: 1949.Jan to 1960.Dec MODEL ESTIMATION/EVALUATION Estimation converged in 6 ARMA iterations, 19 function evaluations. ARIMA Model: (0 1 1)(0 1 1) Nonseasonal differences: 1 Seasonal differences: 1 Standard Parameter Estimate Errors ----------------------------------------------------- Nonseasonal MA Lag 1 0.4018 0.07887 Seasonal MA Lag 12 0.5569 0.07626 Variance 0.13481E-02 ----------------------------------------------------- Likelihood Statistics ------------------------------------------------------------------ Effective number of observations (nefobs) 131 Number of parameters estimated (np) 3 Log likelihood 244.6965 Transformation Adjustment -735.2943 Adjusted Log likelihood (L) -490.5978 AIC 987.1956 AICC (F-corrected-AIC) 987.3845 Hannan Quinn 990.7005 BIC 995.8212 ------------------------------------------------------------------ Roots of ARIMA Model Root Real Imaginary Modulus Frequency ----------------------------------------------------------- Nonseasonal MA Root 1 2.4888 0.0000 2.4888 0.0000 Seasonal MA Root 1 1.7955 0.0000 1.7955 0.0000 ----------------------------------------------------------- Forecasting Airline from 1960.Dec (24 forecasts) Confidence intervals with coverage probability (0.95000) On the Original Scale --------------------------------------- Date Lower Forecast Upper --------------------------------------- 1961.Jan 419.15 450.42 484.03 1961.Feb 391.47 425.72 462.96 1961.Mar 435.92 479.01 526.35 1961.Apr 443.93 492.40 546.17 1961.May 455.02 509.05 569.50 1961.Jun 517.29 583.34 657.84 1961.Jul 589.71 670.01 761.24 1961.Aug 583.00 667.08 763.28 1961.Sep 484.57 558.19 642.99 1961.Oct 428.88 497.21 576.43 1961.Nov 368.52 429.87 501.43 1961.Dec 406.73 477.24 559.98 1962.Jan 415.66 495.93 591.70 1962.Feb 388.71 468.73 565.21 1962.Mar 433.00 527.40 642.38 1962.Apr 440.88 542.15 666.69 1962.May 451.65 560.49 695.55 1962.Jun 513.05 642.28 804.07 1962.Jul 584.32 737.70 931.35 1962.Aug 577.05 734.47 934.84 1962.Sep 479.07 614.58 788.43 1962.Oct 423.49 547.44 707.67 1962.Nov 363.44 473.30 616.38 1962.Dec 400.59 525.46 689.26 ---------------------------------------

regARIMA Model Example

This example uses the "Unit Auto Sales" series of the U.S. Bureau of Labor Statistics. It features in Section 5.4 in the article "New Capabilities and Methods of the X-12-ARIMA Seasonal Adjustment Program" (JBES, 16:127-177). It is example 8 (beauto0.spc) from the set of examples supplied with X-12-ARIMA by the US Census Bureau. To start, load the file beauto.in7 as supplied with X12arima for OxMetrics. The following steps are required to formulate the model:

Model Output

The model output is again available graphically; the next figure gives the seasonally adjusted variable with the original in the first graph, the trend with original in the second, and the irregular in the third graph:

Clicking on the batch editor shows the batch code generated by X12arima for OxMetrics:

series{ file="BeAuto.in7" name=beauto } transform{ function=log } regression{ variables=( const td ao1975.2 ao1988.12 ) user=( "User.Aug" "User.Sep" "User.Oct" "User.Nov" "User.D&J" ) usertype=( ao ao ao ao ao ) file="BeAuto.in7" } x11{ } Apart from the data loading section, this is compatible with the .spc file syntax as used by the command line version of X-12-ARIMA. The batch file, when run from OxMetrics will set the specification, but not do the estimation. This requires an x12run; statement at the end of the batch code.


Batch Usage

The X-12-ARIMA specification syntax is compatible with the OxMetrics batch language. So, if X12ARIMA/OxMetrics is active you can:

The first allows using X12ARIMA/OxMetrics as a .spc file generator for the stand-alone X-12-ARIMA. This will require adding the appropriate series command to read in the data.

Note that only those commands which differ from the default are generated.


Additional Batch Commands

The following commands are specific to X12ARIMA/OxMetrics and are used to supplement the X-12-ARIMA specification statements:


Specification Syntax, Additions and Differences

The file= and name= options are used with specific purpose: the former specifies the OxMetrics database, while the latter lists the selected variables from that database. In the series spec, multiple names can be listed as name=( name1 name2 ).

The model= option in the automdl spec is an addition to avoid the use of an .mdl file. Multiple model= commands may be present to list the models to use in automatic model selection.

Note that only those commands which differ from the default are generated.

Currently, the following subset is implemented:

series{ span=... file=... name=... modelspan=... } transform{ function=... power=... adjust=... } x11{ mode=... sigmalim=... seasonalma=... trendma=... x11easter=... force=... type=... final=... } x11regression{ variables=... user=... usertype=... file=... span=... aictest=... tdprior=... sigma=... outliermethod=... critical=... } regression{ variables=... user=... usertype=... file=... aictest=... augmentusertd=... } arima{ model=... ar=... ma=... } automdl{ mode=... method=... qlim=... fcstlim=... bcstlim=... overdiff=... identify=... outofsample=... model=... } estimate{ maxiter=... tol=... parms=... exact=... outofsample=... } outlier{ types=... method=... critical=... span=... lsrun=... } forecast{ maxlead=... maxback=... probability=... exclude=... }


This file last changed . © JA Doornik