Last updated: 9 January 2008
2007 Washington, DC: Seminars on Stata
2 November 2007
Residence Inn® by Marriott® Arlington Pentagon City
550 Army Navy Drive
Arlington, VA 22202
Proceedings
Statistical computing
Finis Welch
Economist, cofounder and board member
Statistical computing has an interesting history, spanning from the punch
cards to personal computers. This talk will focus on the early developments
in statistical computing and their effect on modern software.
Survival analysis
Roberto G. Gutierrez
Director of statistics
Stata’s capabilities in survival analysis are vast, ranging from simple
Kaplan–Meier curves to complex models of time to failure, such as the Cox
model with group-level random effects. Stata is unique in its ability to
separate the tasks of survival data declaration (time variables, censoring,
truncation, delayed entry, etc.) from the tasks of the actual analysis. The
former are accomplished with stset; the latter, with
other members of the st family of commands,
providing nonparametric, semiparametric (Cox), and fully parametric
analysis.
Longitudinal and panel data
David Drukker
Director of econometrics
This
talk shows how to use Stata to manage and analyze longitudinal/panel data.
The talk covers estimators for the parameters of linear and nonlinear
fixed-effects and random-effects models, as well as dynamic models.
Analyzing complex survey data
Roberto G. Gutierrez
Director of statistics
Most of Stata’s estimation commands (including those for survival
data) are equipped to handle data from complex surveys. That is, estimates
and their standard errors are adjusted for the stratification, clustering
(primary sampling units), and weighted sampling that can occur with survey
data. The adjustment is automatic, provided that the survey aspects of the
data are properly declared using svyset. Learning
to use svyset is therefore the key to making full
use of Stata’s features for survey data.
Extensibility of Stata
William Gould
President and head developer
You can add new features to Stata using the same tools the developers at
StataCorp use, and the result can be indistinguishable from Stata’s
built-in capabilities. Stata provides ways to share and find user-added
features over the web. So even if you never develop new features yourself,
you can easily find and use features added by others. This brief talk gives
an overview of how to do this in Stata.
Time series
David Drukker
Director of econometrics
Stata has a large repertoire of tools for dealing with time series. This
talk shows how to use these Stata tools to manage and analyze time-series
data and discusses methods for univariate and multivariate models for
stationary series and models for vector-cointegrating series.
Multilevel mixed-effects modeling
Roberto G. Gutierrez
Director of statistics
Mixed-effects models contain both fixed and random effects. Fixed effects
are analogous to regression coefficients, whereas random effects are
summarized according to their variance components. Random effects can also
have a nested structure, resulting in a multilevel model. Stata’s
commands for fitting such models are xtmixed for
continuous responses, xtmelogit for binary
responses, and xtmepoisson for count responses.
All three commands share a similar syntax, both for model specification and
for postestimation analysis.
Statistical matrix programming
William Gould
President and head developer
This talk shows how to program statistical methods in Mata, the fast
matrix-programming language in Stata. After demonstrating the simple
methods for putting data into matrices and basic matrix-programming
techniques, the talk illustrates the power of the language by quickly
implementing a Pearson chi-squared goodness-of-fit statistic, a multivariate
regression estimator, and a nonlinear generalized method-of-moments
estimator.
Logistics organizers
Chris Farrar, StataCorp
Gretchen Farrar, StataCorp
