Releases: paulnorthrop/bang
bang: Bayesian Analysis, No Gibbs v1.0.4
bang 1.0.4
Bug fixes and minor improvements
- Fixed 2 \link{} targets with missing (rust) package anchors in Rd files for
hanova1()
andhef()
.
bang: Bayesian Analysis, No Gibbs v1.0.3
bang 1.0.3
Bug fixes
- Fixed issues with the incorrect use of \itemize in some Rd files.
bang: Bayesian Analysis, No Gibbs
bang 1.0.2
Bug fixes and minor improvements
-
In
beta_init_ests()
the initial estimate ofbeta
has been corrected in the casevp > mp * (1 - mp)
, wherevp
andmp
are respectively the sample variance and sample mean of the input probabilities. Thank you to Thomas Richardson for spotting this. -
Create the help file for the package correctly, with alias bang-package.
-
Activated 3rd edition of the
testthat
package
bang: Bayesian Analysis, No Gibbs
bang 1.0.1
Bug fixes and minor improvements
-
pkgdown documentation at https://paulnorthrop.github.io/bang/
-
Added a check that the argument
hpars
tohef()
and tohanova1()
is valid. -
Corrected typos in the vignette Hierarchical 1-way Analysis of Variance. In the 9th line of the Appendix the expression given for ndot was incorrect. It should simply be the sum of the sample sizes in the individual groups. In the subsection "Marginal posterior density of phi" in the Appendix (ci-bi/ai)^2 has been corrected to (ci-bi^2/ai). This is a typo: the ultimate expressions are correct.
-
The summary method for class "hef" is now set up according to Section 8.1 of the R FAQ at (https://cran.r-project.org/doc/FAQ/R-FAQ.html).
-
Corrected an error that meant that the S3 plot method plot.hef() was not resetting graphical parameters correctly on exit.
bang: Bayesian Analysis, No Gibbs
Bayesian Analysis, No Gibbs
What does bang do?
Provides functions for the Bayesian analysis of some simple commonly-used models, without using Markov Chain Monte Carlo (MCMC) methods such as Gibbs sampling. The 'rust' package https://cran.r-project.org/package=rust is used to simulate a random sample from the required posterior distribution, using the ratio-of-uniforms method. Currently three conjugate hierarchical models are available: beta-binomial, gamma-Poisson and a 1-way Analysis of Variance (ANOVA). Advantages of the ratio-of-uniforms method over MCMC in this context are that the user is not required to set tuning parameters nor to monitor convergence and a random posterior sample is produced. See the 'bang' website for more information, documentation and examples.
A simple example
The hef
function samples from the posterior distribution of the parameters of certain hierarchical exponential family models. The following code performs essentially the same analysis of the rat tumor data using a beta-binomial hierarchical model that appears in Section 5.3 of Gelman, A., Carlin, J. B., Stern, H. S. Dunson, D. B., Vehtari, A. and Rubin, D. B. (2014) Bayesian Data Analysis. Chapman & Hall / CRC. http://www.stat.columbia.edu/~gelman/book.
rat_res <- hef(model = "beta_binom", data = rat)
plot(rat_res)
Installation
To get the current released version from CRAN:
install.packages("bang")