Bayesian Estimation Supersedes the t-Test

An alternative to t-tests, producing posterior estimates for group means and standard deviations and their differences and effect sizes. It implements the method of Kruschke (2013) Bayesian estimation supersedes the t test. Journal of Experimental Psychology: General, 142(2):573-603 .


CHANGES in v.0.5.1 (2018-04-19)

  • Fixed issue with position of text on density plots.

  • plotAll and plotPostPred modified to show the data with an 'ordinary' histogram.

  • Warning if adaptation is not adequate (instead of "NOTE: Stopping adaptation").

CHANGES in v.0.5.0 (2017-05-28)

  • 'BESTmcmc' uses 'rjags' directly, instead of 'jagsUI' wrappers. This resolves 'set.seed' issues, but values returned will not be the same as with previous versions.

  • Function 'hdi' removed; imports HDInterval::hdi instead.

CHANGES in v.0.4.0 (2015-12-28)

  • BESTmcmc now allows normality parameter, nu < 1 (See John Kruschke's blog post 2 Dec 2015).

  • Vignette updated to use new prior specification.

CHANGES in v.0.3.0 (2015-06-29)

  • BESTmcmc 'showPriors' argument changed to 'doPriorsOnly'.

  • Function postPriorOverlap added.

CHANGES in v.0.2.6 (2015-05-09)

  • BESTmcmc gains a 'showPriors' argument.

  • Removed constraints on small samples in BESTmcmc (n=1 is ok if priors are specified).

  • makeData throws an error if the number of outliers is 1.

CHANGES in v.0.2.5 (2015-05-07)

  • Run chains in parallel using jagsUI (>= 1.3.6) instead of rjags.

CHANGES in v.0.2.4 (2015-05-03)

  • Function BESTmcmc now allows user to specify priors.

CHANGES in v.0.2.3 (2014-11-17)

  • Added sanity checks to BESTmcmc.

CHANGES in v.0.2.2 (2014-06-06)

  • hdi gains a density method, which can handle discontinuous HDIs. This is used for the posterior density plots when showCurve = TRUE.

  • Enhancements to plots: o credMass = NULL suppresses plotting of the credible interval. o plotPost, the plot.BEST method, and plotAreaInROPE now make full use of the ... argument for additional graphical parameters. o default xlim for posterior density plots is now based on 99% HDI (not range). o plotPost and plot.BEST return an object of class 'histogram' invisibly. o plotPost deals properly with a vector of integers.

CHANGES in v.0.2.0 (2013-08-23)

  • Added function plotAreaInROPE.

  • In BESTpower, call to x11() (which is platform specific) changed to; in Rstudio, the plot is shown on active device.

  • BESTmcmc and BESTpower gain 'rnd.seed' arguments for reproducible output; testthat test files for BESTmcmc and BESTpower added.

  • makeData gains a showPlot argument.

  • Fixed bug in plot.BEST: the ... argument was not being passed to plotPost.

  • BESTmcmc has 'verbose' argument; setting verbose = FALSE (or 0) suppresses output to the Console (wish of Ben Marwick).

  • BESTpower also has 'verbose' argument; here verbose = 0 (or FALSE) suppresses all output, verbose = 1 (or TRUE) provides just a progress bar; verbose = 2 (the default) displays full details.

Reference manual

It appears you don't have a PDF plugin for this browser. You can click here to download the reference manual.


0.5.4 by Mike Meredith, 11 days ago

Report a bug at

Browse source code at

Authors: John Kruschke [aut] , Mike Meredith [aut, cre]

Documentation:   PDF Manual  

Task views: Bayesian Inference

GPL (>= 3) license

Imports coda, rjags

Depends on HDInterval

See at CRAN