NEWS.md

In GabrielHoffman/mvBIC: Multivariate Information Criteria to identify important predictors in high dimensional multivariate regression

mvIC 1.6.3

• August 30, 2022
• update R and variancePartition dependency versions

mvIC 1.6.2

• August 22, 2022
• remove dependency on PRIMME package

mvIC 1.6.1

July 26, 2021

• fix bug in calculating effective degrees of freedom:
• report m = mean edf per response
• for fixed effect, now report m instead of m + 1
• Add more unit tests
• in pcTransform() varFrac is 1 but default
• enable pca by default, and compute it only once in mvForwardStepwise()

mvIC 1.6.0

Major changes

July 13, 2021

• add fastApprox argument that is exact for fixed effects and approximate if random effects are included
• include pcTransform() function
  • this can be 100x faster for large datasets
• better estimate of lambda and logML
  • use lambda = 0.01 by default
• defaults are new criterion = "BIC", shrink.method = "EB", nparamsMethod = "edf"
• remove scaling of responses to allow use of pcTransform() on responses
• Using effect degrees of freedom is faster since it is built into dream()
  • fixed bugs handling edf for each regression
  • remove redundant calculations
• Add more unit tests

mvIC 1.5.1

October 26, 2020 - enforce compatability with variancePartition >= 1.19.20

mvIC 1.5.0

May 5, 2020 - add back gdf

mvIC 1.4.0

May 1, 2020 - Add bestSubsetSearch() - Add estimateMVN_EB for Empirical Bayes estimation from Multivariate Normal Inverse Wishart model (MNIWEB)

mvIC 1.3.0

May 1, 2020 - add rough code for empirical Bayes covariance estimation

mvIC 1.1.4

April 27th, 2020 - Add other selection criteria

mvIC 1.1.4

April 26th, 2020 - Add fast shrinkage of eigen values for Touloumis_equal

v1.1.3

April 26th, 2020 - add generalized degrees of freedom for covariance estimate

v1.1.1

April 16th, 2020 - Improve documentation - add naive BIC method - add print and show methods

v1.1

April 15th, 2020 - define number of parameters for BIC as nparamsMethod = c("edf", "countLevels", "lme4") - set default to effective degrees of freedom: "edf" this reduce the penalty for random effects compared to counting the levels - set stopping criterion to delta of 10

GabrielHoffman/mvBIC documentation built on Sept. 5, 2022, 1:57 a.m.