BAYESIAN ANALYSIS COMMANDS
Interest and need of Bayesian analysis is on the rise and come from users from different disciplines, primarily from health/medicine, economics and other social sciences. Bayesian analysis deals with research questions with unknown parameters of statistical models by resorting to probability statements with an assumption that all the parameters of the model are subject to some prior knowledge.
Stata 14 introduces a new command bayesmh which allows users to fit models using an adaptive Metropolis-Hastings algorithm or a full Gibbs algorithm (or a combination of both). The command bayesmh can actually support univariate, multivariate, and multiple-equations, both linear and nonlinear. Stata 14 also introduces bayesgraph and bayesstats ess which allow users to visually check for convergence and Markov chain Monte Carlo efficiency.
In Stata 14 users can take advantage of 12 built-in likelihood models for all outcomes (continuous, binary, ordinal and count), write their own likelihood models while still using the 22 built-in priors and the postestimation features. Stata 14 also adds an all-new 261 page Stata Bayesian Analysis Reference Manual
.
Return to Stata 14 New Features
Stata 14 introduces a new command bayesmh which allows users to fit models using an adaptive Metropolis-Hastings algorithm or a full Gibbs algorithm (or a combination of both). The command bayesmh can actually support univariate, multivariate, and multiple-equations, both linear and nonlinear. Stata 14 also introduces bayesgraph and bayesstats ess which allow users to visually check for convergence and Markov chain Monte Carlo efficiency.
In Stata 14 users can take advantage of 12 built-in likelihood models for all outcomes (continuous, binary, ordinal and count), write their own likelihood models while still using the 22 built-in priors and the postestimation features. Stata 14 also adds an all-new 261 page Stata Bayesian Analysis Reference Manual
| A selected list of available models and features in Stata 14 for Bayesian analysis: |
|
| Available Models | Available Features |
| Univariate linear regression models | 12 built-in likelihood models |
| Univariate GLM | 22 built-in prior distributions |
| Multivariate normal linear models | Continuous, binary, ordered, and count outcomes |
| Univariate nonlinear regression models | Continuous and discrete prior distributions |
| Univariate generalized nonlinear models | Adaptive Metropolis-Hastings and Gibbs methods |
| Multivariate normal nonlinear models | Facilities for writing your own Bayesian models |
| Multiple-equation models | Graphical convergence diagnostics Checks for Markov chain Monte Marlo (MCMC) efficiency Posterior means, medians, credible intervals, and more Bayesian information criteria and Bayes factors Interval hypothesis testing Model posterior probabilities |
Return to Stata 14 New Features