NEWS
bgms 0.1.4.2
Fixes a bug with the adaptation of the proposal variances.
bgms 0.1.4.1 (2024-11-12)
This is a minor release that adds some documentation and bug fixes.
bgms 0.1.4 (2024-10-20)
New features
- Comparing the category threshold and pairwise interaction parameters in two independent samples with bgmCompare().
- The Stochastic Block model is a new prior option for the network structure in bgm().
Other changes
- Exported extractor functions to extract results from bgm objects in a safe way.
- Changed the maximum standard deviation of the adaptive proposal from 2 to 20.
- Some small bug fixes.
bgms 0.1.3 (2024-02-25)
New features
- Added support for Bayesian estimation without edge selection to bgm().
- Added support for simulating data from a (mixed) binary, ordinal, and Blume-Capel MRF to mrfSampler()
- Added support for analyzing (mixed) binary, ordinal, and Blume-Capel variables to bgm()
User level changes
- Removed support of optimization based functions, mple(), mppe(), and bgm.em()
- Removed support for the Unit-Information prior from bgm()
- Removed support to do non-adaptive Metropolis from bgm()
- Reduced file size when saving raw MCMC samples
bgms 0.1.2 (2023-10-13)
This is a minor release that adds some bug fixes.
bgms 0.1.1 (2023-09-01)
This is a minor release adding some new features and fixing some minor bugs.
New features
- Missing data imputation for the bgm function. See the
na.action
option.
- Prior distributions for the network structure in the bgm function. See the
edge_prior
option.
- Adaptive Metropolis as an alternative to the current random walk Metropolis algorithm in the bgm function. See the
adaptive
option.
User level changes
- Changed the default specification of the interaction prior from UnitInfo to Cauchy. See the
interaction_prior
option.
- Changed the default threshold hyperparameter specification from 1.0 to 0.5. See the
threshold_alpha
and threshold_beta
options.
- Analysis output now uses the column names of the data.