U.S. flag

An official website of the United States government

Skip Header

Notes on X-13ARIMA-SEATS Version 1.1, Build 39

Build 39 of Version 1.1

This version of the program gives slightly different seasonal adjustments from Build 26 (the last release) for the sample data sets that I have used for previous testing. This is due to the standardization of pi throughout the software, making each instance of π (pi) have the same number of significant digits. The algorithm we use to generate end weights for the Henderson filter uses π, as do our spectral routines.

The largest difference in the seasonal adjustment encountered in the practice set is 4.432105e-09.

There were no changes in peak detections in the practice set from Build 26 to Build 39.

Note: this version should be used with Version 2.5 of Win X-13.

New options

  • Added easter[0] regressor to available regressors.
  • Added testalleaster argument to regression spec to test all Easter regressors in the AICC testing procedure.
  • Added qcheck argument to spectrum spec to add test of seasonality for quarterly version of the original series and seasonally adjusted series of monthly series.
  • Added qlimit argument to check spec to set p-level of test for ACF lags in .udg and log output.
  • Added runtime argument -t to store timing information in the .udg file.

Changes to the program output

  • Do not print header for spectral output in log file unless the user saves peaks info to log file.
  • Corrected x11regression AIC test table in log output of HTML output version.
  • Reduced sliding spans printout and don't save diagnostics if differences are analyzed.
  • Added index entry in HTML output for outlier activity with the history analysis.
  • Revised output for outlier activity of history analysis.
  • Restricted output for model estimation errors during history and sliding spans analysis to error output.
  • Revised HTML output of model summaries.
  • Indented HTML output for selected sections of the sliding spans header output.
  • Added heading for spectral peak output to log output.
  • Printed model error messages in correct format for HTML output.
  • Changed output of model estimation iterations to handle large numbers.
  • Changed output of histogram to allow printing of large numbers.
  • Changed output of sliding spans summary tables to allow printing of large numbers.
  • Added space after HTML output of irregular regression trading day table.
  • Added spaces to HTML output of Tukey peak analysis.
  • Added space after HTML output of trading day weight table.
  • Corrected error message for model root analysis.
  • Revised format for printing out Geary's a.
  • Ensured that error message for squared ACF is always printed to the output file.

Changes to .udg output

  • Added initial difference results from automatic model identification procedure to the .udg file.
  • Added number of Easter regressors in AIC test to .udg file.
  • Differentiated between lom, loq and lpyear regressors in listing for aictest in .udg file.
  • Changed x12model choice for savelog argument in seats spec to x13model (note - x12model still a valid choice, though currently undocumented).
  • Removed double ':' from .udg file.
  • Added number of observations to .udg output.
  • Added number of observations in span to .udg output.
  • Added number of observations in outlier span to .udg output.
  • Added number of effective observations in outlier span to .udg output.
  • Added "SEATS seasonal adjustment" as choice of samode in .udg output.
  • Added 5 best models from automdl with bic to .udg output.
  • Added listing of significant ACF and PACF lags in .udg output.
  • Changed the output for the aictest.e.window entry in the .udg output if an Easter regressor is not selected using aictest=easter from 0 to -99999.

Internal program changes

  • Standardized all instances of π (pi) in the software to have the same number of significant digits.
  • Added specific user-defined regressor types for trading day, length of month, length of quarter, AO, LS, SO, constant, etc.
  • Turned off sliding spans analysis for seasonal adjustments and trading day when span argument is specified.
  • Added transitory regressors to regression type printed out in regression effects matrix.
  • Initialized original series variable.
  • Added check for whether both seats and x11 specs are included in the spec file.
  • Regenerated regression matrix before printing out or saving regression effects.
  • Disabled the opening of files in the SEATS part of the program when sliding spans and/or history analysis is done.
  • Reduced the number of choices for usertype in the x11regression spec.
  • Added a variable for the number of Easter regressors.

Fixes to defects in previous versions

  • Corrected arguments for Tukey spectrum peak probabilities for indirect seasonal adjustment.
  • Ensured the input format x13save is processed correctly by the input routines.
  • Corrected problem with storing title when the length of the series name is longer than 80 characters.
  • Computed correct difference between start of original series and start of model span in the sliding spans model checking routines.
  • Corrected length of string written to for non-log transformations.
  • Fixed chi-squared test with fixed regressors, user defined holiday and trading day.
  • Revised routines for trading day and seasonal F-test to work better with user-defined trading day regressors.
  • Corrected degrees for freedom for trading day and seasonal F-tests.
  • Corrected peak identification for final spectral frequency.
  • Recoded loop used in outlier identification to avoid an array out of bounds error.
  • Recoded line printer plots to ensure observations are not incorrectly determined to be missing.

We are interested in your feedback. Please email your questions and comments to x12a@census.gov.

Page Last Revised - October 8, 2021
Is this page helpful?
Thumbs Up Image Yes Thumbs Down Image No
255 characters maximum 255 characters maximum reached
Thank you for your feedback.
Comments or suggestions?


Back to Header