About BayES


Supported Models


Development Plans



BayES >> Bayesian Econometrics Software

BayES is a software package designed for performing Bayesian inference in some popular econometric models using Markov Chain Monte Carlo (MCMC) techniques. Bayesian inference traditionally requires technical skills and a lot of effort from the part of the researcher, both in terms of mathematical derivations and computer programming. BayES provides canned procedures for performing Bayesian inference on specific models, thus avoiding the time-consuming process of deriving complete/full conditionals and coding the samplers.

BayES is menu driven, providing an intuitive user interface that enables first-time users to run a model in a matter of minutes (see also the video tutorials). On the other hand, it also features a rich scripting language that allows fast and efficient communication with the program, while facilitating reproducibility of the analysis. Users familiar with other popular matrix languages will find the transition to BayES' scripting language a breeze.

Modern Bayesian inference is based on very computationally intensive techniques. On the other hand, MCMC methods are "embarrassingly parallel". BayES supports multi-threading, thus making estimation much faster. Future development plans include taking advantage of Graphical Processing Units (GPUs), which are present in most personal computers. Such a development could increase speed by orders of magnitude.

BayES is not simply another econometrics software package. Its emphasis is in models that are hard or impossible to estimate using classical (frequentist) inference and, thus, it is not a substitute for mainstream and well-established econometric software packages. On the other hand, BayES is not competing with general or statistical scripting languages either. BayES is designed for the user who wants to perform Bayesian inference in a computationally involved problem, but who does not want to learn a new programming language for doing so.

BayES is built to perform some very specific tasks and it recognizes that there could be other software packages available that are more appropriate for other types of tasks. Instead of trying to do everything by itself, BayES provides interfaces to these packages to facilitate communication. This is achieved by means of software-specific interface functions, which allow the user to pass data and native code to, execute scripts in and retrieve results from each package, without ever leaving the BayES environment.

Currently Supported Models

BayES is still at an early development stage. The models that are supported are:

  • linear models for cross-sectional and panel data:
    • simple linear model
    • random-effects model
    • random-coefficients models
    • latent-class models
  • stochastic frontier models (production and cost specifications):
    • simple stochastic frontiers
    • inefficiency-effects models
    • random-effects stochastic frontiers
    • random-coefficients stochastic frontiers
    • latent-class models
    • dynamic stochastic frontiers, with and without random effects
  • discrete-choice models:
    • simple Probit and Logit models
    • random-effects Probit and Logit models
    • multivariate Probit model
    • multinomial and conditional Probit and Logit models
  • models where the dependent variable has an ordinal interpretation:
    • simple ordered Probit and Logit models
  • models for count data:
    • simple Poisson and negative-Binomial models
  • models for censored and truncated variables:
    • simple type I Tobit models
    • simple type II Tobit models
  • models for systems of equations:
    • simple Seemingly Unrelated Regressions (SUR) models
  • time-series models:
    • Vector Autoregressive (VAR) models

See the documentation section of this website for more details.

Future Development Plans

The following models are planned for future inclusion in BayES:

  • linear models:
    • Markov-switching models for time-series and panel data
    • State-Space models for time-series and panel data
  • stochastic frontier models:
    • Markov-switching models for panel data
  • discrete-choice models:
    • random-coefficients Probit and Logit models
    • random-effects multinomial and conditional Probit and Logit models
    • nested Logit models
  • models for censored and truncated variables:
    • hurdle models
    • models for truncated regression
  • models for count data
    • zero-inflated Poisson models
    • Poisson and negative-Binomial models with random effects
  • time-series models:
    • ARMA and ARIMA models
    • univariate and multivariate ARCH and GARCH models
    • Vector Error Correction (VEC) models
    • univariate and multivariate stochastic-volatility models

Development is currently focused on the implementation of latent-class models. If you would like to see particular models or features in BayES or if you have a preference on which of the models presented above should be implemented first, consider expressing your views by taking a short survey.



Under the Hood

Under the hood





The current version of BayES is provided free of charge. However, BayES is not an open-source project and future versions may be provided for a fee. Please read the license agreement before installing and start using BayES. You will need to accept the terms and conditions of the license agreement before you are able to use BayES.

A Peek Under the Hood

BayES is written entirely in c++. The graphical user interface is built using wxWidgets. BayES proudly uses Eigen for its matrix calculations and parts of the Boost c++ libraries. ANTLR is used to parse BayES' language and PLplot is used for plotting.

Estimation is performed using Gibbs samplers, whenever possible. If the full conditional for a parameter does not belong to any known family, then BayES uses case-specific Metropolis-Hastings updates.

Installers for the Latest Release of BayES

BayES runs on 32-bit and 64-bit Microsoft® Windows® systems (Windows 7, 8, and 10), on 32-bit and 64-bit Linux systems and on 64-bit macOS systems

The following table contains links to the latest-release installers. Previous releases can be found in the downloads section of the website. Step-by-step instructions for installing BayES on Microsoft® Windows® or Linux systems are provided in Appendix A of the documentation.

PIC If you download and use BayES, please consider providing feedback. Positive feedback is always welcome and constructive criticism much appreciated.

You can provide feedback by taking a short survey, by email or by commenting on the project's page on ResearchGate.
Operating System / Architecture Version Installation File Release Date
Microsoft Windows / 64bit 2.5 BayES_Win64.exe February 2020
Microsoft Windows / 32bit 2.5 BayES_Win32.exe February 2020
Linux / 64bit 2.5 BayES_Linux64.bsx February 2020
Linux / 32bit 2.5 BayES_Linux32.bsx February 2020
macOS / 64bit 2.5 BayES_macOS64.bsx February 2020
Share this content:
Facebook Twitter LinkedIn Email
© 2016–20 Grigorios Emvalomatis