[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]
st: Stata 10 announcement
There is a long tradition of first announcing new releases of Stata on
Statalist. Here it is:
Stata 10 ships June 25.
Orders are now being accepted by Stata and our distributors.
800-782-8272 (800-STATAPC, USA)
(+1) 979-696-4600 (Worldwide, voice)
(+1) 979-696-4601 (Worldwide, FAX)
More information is available at
A new feature that will interest everyone is
o Stata's new WYSIWYG Graph Editor. Point. Click. Edit.
Marketingwise, the Graph Editor may eclipse the other new features, but
we hope not. Also of interest to everyone will be
o Date/time variables. Stata already had date variables; now it
has time variables (and time-and-date variables, too). Time is
recorded with up to millisecond resolution, which will be
important to some of you. Also, Stata's time variables will
work with or without leap-second adjustment. You may never
have heard about leap seconds, but there are clocks that are
adjusted for them, and clocks that are not; in computer-
generated datasets, time stamps come both ways. Stata can work
with either one.
o Estimation results may be saved to disk. You type -estimates
save <filename>- to save results and -estimates use <filename>-
to use them. When you use results, everything is just as it
was, meaning that you can use any postestimation command, such
as -test-, -predict-, and -nlcom-.
Statisticswise, Stata 10 includes
o Exact logistic and Poisson regression. These new estimators
will be important to biostatisticians, those in related fields,
and anyone who works with small datasets.
o Multilevel, nested, hierarchical, and mixed-effects logistic
and Poisson regression. This is basically the equivalent of
Stata 9's -xtmixed-, but for binary and count responses.
o More estimators for dynamic panel-data models, including the
Arellano-Bover/Blundell-Bond system estimator and its
generalizations. Economists, take note.
o LIML and GMM regression in addition to 2SLS, including a test
for weak instruments. (LIML stands for Limited Information
Maximum Likelihood and GMM stands for Generalized Method of
o Nonlinear, seemingly unrelated regression.
o More survey estimators, including Cox proportional hazards
regression for survey data. As for the rest, Stata 10 includes
27 more survey estimator. Stata 9 had 21 estimators. We have
more than doubled what is available.
o Discriminant analysis, including linear, quadratic, and
o Multiple correspondence analysis (MCA) and joint correspondence
o Sample-size and power calculations for survival studies.
There's lots more -- about 21 pages -- visit
http://www.stata.com/stata10/ for the complete list.
1. You know that we recently released Stata/MP, Stata for
multiprocessor computers, including dual-core CPUs. Two
CPUs or cores do not double performance, but Stata/MP
achieves 72% efficiency; see http://www.stata.com/statamp/.
For Stata 10, even more is parallelized. If you don't
have Stata/MP, you can upgrade to it when upgrading to Stata 10.
2. Because of the new time variables, Stata's .dta dataset format
has changed. This happens periodically and, because Stata
always reads old-format datasets, most users don't care or even
notice. It does mean, however, that if you use Stat/Transfer,
you will want to upgrade to Stat/Transfer 9 [sic]. Yes, it's
confusing: Stat/Transfer 9 goes with Stata 10.
-- Bill (for the entire Stata 10 development team)
* For searches and help try: