bayesm

v3.1-7:

• changed out deprecated function for a minimum based on suggestions from Dirk E.

v3.1-6:

• changed out deprecated function for initializing a vector based on suggestions from Dirk E.

v3.1-5:

• rhierMnlRwMixture - fixed indexing in warning re large/small values of beta_pooled.
• fixed minor comparison error- complained about by CLANG.
• removed bad URL links to old BSM book site – google “Bayesian Statistics and Marketing” instead
• changed sprintf to snprintf in clusterMix_rccp_loop

v3.1-4:

• rordprobitGibbs fixed setting of epsilon in subroutine lldstar

v3.1-3:

• rhierMnlRwMixture updates were left out of v3.1.-2 by mistake. Added. Also, changed to Nelder-Mead method for finding pooled constrained optimum (used for Metropolis tuning only). Added warning regarding large values of constrained pooled coefficients. (coefficients larger than 10 in absolute value – exp(10) is too large and exp(-10) is too small)

• plot.bayesm.hcoef method changed to improve labelling of plots.

v3.1-2:

• rhierMnlRwMixture error in Hessian computation for Metropolis tuning corrected. Default initial parameter values used for tuning optimization also changed.

• all plot.bayesm.<> and summary.bayesm.<> methods have added checks to insure that burnin is set no larger than number of submitted draws.

v3.1-1:

• rbrobitGibbs error in trunNorm_vec corrected

• plot.bayesm.hcoef (method for plotting hierarchical model coefficients) bug in boxplots corrected (median added)

• ghkvec documentation improved wrt to above parameter

v3.1-0:

• rhierMnlRwMixture now accepts sign constraints on beta_i parameters

• rivDP error (\(w\) variables) corrected. Error was introduced in v3.0-0

• llmnp error (involving ghk_vec.C) corrected. Error was introduced in v3.0-0

• camera dataset added

• rnegbinRw and rhierNegbinRw now accept Mcmc$alpha initial value

• rhier* functions now check input types (e.g. rhierMnlRwMixture’s input lgtdata[[1]]$X must be a matrix)

• rmnpGibbs, rmvpGibbs, and rbprobitGibbs have an updated utility routine that improves accuracy of draws from a truncated normal distribution when the truncation points are extreme. rordprobitGibbs and rtrun will be similarly updated in a future version of bayesm.

• rhierBinLogit has been deprecated; please use rhierMnlRwMixture instead

• “bayesm Overview” vignette added

• “Constrained Multinomial Logit” vignette added

• Numerous edits to .Rd help files

• Utility function fsh no longer exported

v3.0-0:

• Most MCMC routines for posterior sampling functions moved to C++ via Rcpp and RcppArmadillo