Bayesian Model Averaging Library

Bayesian model averaging for linear models with a wide choice of (customizable) priors. Built-in priors include coefficient priors (fixed, flexible and hyper-g priors), 5 kinds of model priors, moreover model sampling by enumeration or various MCMC approaches. Post-processing functions allow for inferring posterior inclusion and model probabilities, various moments, coefficient and predictive densities. Plotting functions available for posterior model size, MCMC convergence, predictive and coefficient densities, best models representation, BMA comparison.


News

############################################################### Version 0.3.4 (build 2015-11-13)

Maintenance release to fix standard changes regarding S3 and S4 methods in R version 3 Two arguments added to function bms:

  • data: function bms is now fully compatible with the formula and data arguments familar from function ' lm'. see #' attitude' example in help(bms)
  • randomizeTimer: for geeks, the call to set.seed can be turned off in order to replicate MCMC runs

############################################################### Version 0.3.2 (build 2013-11-17)

Maintenance release to cope with changes in R version 3 compared to version 2

############################################################### Version 0.3.1 (build 2012-08-28)

Maintenance release to fix inconsistencies in package version 0.3.0 (Warnings in package checks) under new R versions >=2.14

############################################################### Version 0.3.0 (build 2011-05-04) (changes compared to version 0.2.5)

  • USER-VISIBLE CHANGES TO EXISTING FUNCTIONS: ------------------------

    • Changed behavior of pmp.bma (see help(pmp.bma))
  • NEW FEATURES: ------------------------

    • bms now can hold a set of fixed regressors (argument fixed.reg)
    • New functions pred.density, plot.pred.density, print.pred.density, predict.bma, predict.zlm
    • New functions quantile.density, quantile.pred.density, quantile.coef.density
    • New function pmpmodel
    • New functions post.var and post.pr2
    • New functions model.frame.bma, model.frame.zlm, deviance.bma, deviance.zlm, variable.names.bma, variable.names.zlm, logLik.zlm, vcov.zlm
    • bms can now accommodate user-defined coefficient (g) and model-priors, as well as samplers
  • BUG FIXES: ------------------------

    • bms: + bug with accept.candi = NA fixed + problems with environment in non-standard settings fixed (in internal function .construct.arglist) + mprior =="custompip" is now working properly (fixed in internal function .mprior.pip.init) + fixed problem with sys.call in non-standard calling environments +
    • plotComp: +also works for single bma object now + fixed display when variable names are too long + fixed many problems with graphical parameters - completely redone
    • topmod: + bug when likelihood==-Inf, fixed (in internal function .top10) + initializing with bbeta==FALSE now works properly (fixed in .top10)
    • density.bma: bug with par and plot.new fixed when argument addons contains "p"
    • image.bma: fixed display problems when variable names too long
    • plotConv (resp. plot.bma): fixed warning 'In min(x) : no non-missing arguments to min; returning Inf'
    • plotModelsize: + bugfixes for parameter ksubset + bugfixes for graphical parameters + can now cope with case when model prior is not available
  • CHANGES TO INTERNAL OBJECTS: ------------------------------

    • New classes: gprior, mprior, pred.density, implicit class coef.density
    • Streamlining and customizability of gprior required changes to following functions: bms, zlm, .choose.gprior, gdensity, .topmod.as.bbetaT New functions gprior.constantg.init, gprior.eblocal.init and .gprior.hyperg.init now provide for generic handling of g-priors Old functions lprob.*.init are still available for compatibility
    • Streamlining and customizability of mprior required changes to following functions: .choose.mprior, and all .mprior.*.init functions
    • Streamlining and customizability of sampling functions led to streamlining of: .iterenum, .fls.samp, .rev.jump, .enum_startend
    • Introduction of bms argument fixed.reg is handled generically by functions
      .fixedset.mprior, .fixedset.sampler, .starter, .choose.mprior, bms
    • Speeding-up of syntax (replacing $ with [[, directly calling internal functions) yields 5% speed gain in bms vs version 0.2.5
    • .construct.arglist has been vastly changed

############################################################### Version 0.2.5 (build 2010-07-30)

  • NEW FEATURES

None

  • BUG FIXES

  • fixed a bug affecting coefficient appearing

Reference manual

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

install.packages("BMS")

0.3.4 by Stefan Zeugner, 5 years ago


http://bms.zeugner.eu/


Browse source code at https://github.com/cran/BMS


Authors: Martin Feldkircher and Stefan Zeugner


Documentation:   PDF Manual  


Task views: Bayesian Inference, Econometrics


Artistic-2.0 license


Imports graphics, methods, stats


Imported by TBSSurvival, conting, mvMonitoring.

Suggested by highfrequency.


See at CRAN