Our Summary of Stata 14
StataCorp just released the 14th version of its statistical software Stata. Stata 14 offers a combination of new commands and features while also extending existing commands from Stata 13. Stata 14 also expands its reach across borders with Spanish and Japanese translation of its menus and dialogs. With new and expanded commands, new and updated reference manuals, support for World Health Organisation’s ICD-10 codes, ability to read Unicode UTF-8, a theoretical upper limit of more than 281 trillion observations, and language adaption Stata 14 better serves all the different academic disciplines going beyond economists and econometricians reaching out to health researchers, education specialists, psychologists, psychometricians, epidemiologists, biostatisticians, engineers, and political scientists.
Among the new commands in Stata 14 and following up on the rise in popularity of Bayesian Analysis, Stata 14 offers new command bayesmh which allows users to fit models using an adaptive Metropolis-Hastings algorithm or a full Gibbs algorithm. The new command supports univariate, multivariate and multiple-equations, both linear and nonlinear. Stata 14 also offers 12 built-in likelihood models and 22 built-in prior with facilities for writing your own Bayesian models. Read more here about Stata 14’s Bayesian Analysis capabilities.
In addition to Bayesian Analysis, Stata 14 now allows users to fit item response theory (IRT) models for binary, ordinal or categorical items. To accommodate different disciplines, Stata 14 provides the ability to customise the output whether by discrimination, difficulty or item. Further, Stats 14 offers a full range of postestimation graphs to assess both the item and the test. IRT models in Stata 14 are discussed further in an all-new reference manual. Read more here about Stata 14’s Item Response Theory capabilities.
In its expanding support to multilevel survival models, parametric mixed-effects survival models, with either single or multiple-record st data, can now be fit in Stata 14 using the new command mestreg. The new command supports a variety of distributions and parameterisation. Stata 14 adds a new option dfmethod to the existing command mixed which allow users to implement different adjustment methods for small-sample inference for computing degrees of freedom. This expansion goes beyond cross sectional data and reaches panel data estimators with support to random-effects estimators for panel data on survival. Finally, Stata 14 replaces stpower with the new command power in which users can obtain custom tables and automatic graphs of power curves.
In terms of extending existing commands and in the area of treatment effects, the existing command teffects now gets support from a new command stteffects which now extends teffects to allow users to model a combination of the outcome, treatment assignment and censoring. The new command offers different options for estimating treatment effects. Further, with Stata 14 users can easily deal with endogenous treatments where the treatment assignment is correlated with the outcome using the new command eteffects.
Stata 14 has added to its existing set of time series commands by introducing mswitch to fit Markov-switching models using either, an autoregressive (AR) or dynamic regression (DR) processes. The new command is supported with new postestimation commands to test for structural breaks with single unknown date, single known date and multiple known dates.
Timberlake can help you find out more about the new features of Stata 14 through free onsite / online demo of Stata’s new capabilities. You can also check our extensive list of Stata public attendance training courses.
Dr. George Naufal, Technical Director, Timberlake
13 April 2015