Stata Technical Bulletin
25
(Update)
sts7.3 A library of time series programs for Stata
Sean Becketti, Stata Technical Bulletin, FAX 914-533-2902
In sts7, a library of time series programs for Stata was introduced (Becketti 1994). That insert described an approach to
time series analysis that builds on Stata’s core commands and on its extensibility. The insert also cataloged the programs in the
time series library.
This update describes changes and additions to the time series library. An updated catalog of programs is also included. The
updated library is available on the STB diskette. This update will be repeated in each issue of the STB. Consult the original insert
for a general discussion of Stata’s approach to time series analysis. As always, I actively solicit your comments, complaints, and
suggestions.
New features
New date-handling commands: Alan Riley’s new date-handling commands (described in dm20 earlier in this issue) have been
included in the time series library. They are listed in Table 1.
A catalog of programs
Table 1 lists the user-level programs in the time series library. Each program’s status is indicated by a letter grade. An ‘A’
indicates a program that is safe for general use. An ‘A’ program has been documented—in its current form—in the STB and
follows all Stata guidelines for an estimation command, where relevant (see [4] estimate). A ‘B’ program produces accurate
results, but either is not fully documented, not completely compatible with the standard time series syntax adopted in the library,
or not in conformance with the guidelines for an estimation command. Most ‘B’ programs receive that grade because they have
been revised significantly since they were last documented. A ‘C’ program is incomplete in significant ways but can be used
safely by an advanced Stata user. A ‘D’ program has serious deficiencies, however its code may provide a useful model to
advanced Stata users writing their own time series programs. An ‘O’ program is obsolete, that is, it has been superseded by a
newer program. An ‘O’ program is retained if it is still called by one or two user-level programs. There are currently no ‘D’ or
‘O’ programs.
Utilities for time series analysis
Writing programs for time series analysis presents a variety of challenges. In developing this library of programs, I had to
write a pool of utility programs to interpret the time series options, to generate lags, to manipulate the list of variables in a lag
polynomial, and so on. I recommend that you familiarize yourself with these utilities, if you wish to write your own time series
programs. A list of some of the most frequently used utility programs appears in Table 2 below.
Future developments and call for comments
This library of time series programs is under constant revision and extension. Projects under development include programs
to estimate rolling regressions, to estimate vector autoregressions, and to perform maximum-likelihood tests for cointegration.
Older programs are being revised to bring them up to Stata’s standards for estimation programs. A disadvantage of these constant
revisions is the likelihood of inadvertently introducing errors into the programs. The advantage of constant revision is the ease
and rapidity of fixing these errors and the steady increase in Stata’s time series capabilities. I encourage you to alert me to any
errors or inconveniences you find.
If you find an error in any of these commands, I will attempt to correct it by the next issue of the STB. To speed the process,
please send me a diskette containing a do-file that replicates the error. Debugging software is similar to auto mechanics: if I
can’t reproduce the problem, I can’t fix it.