NEWS.md

In MixMatrix: Classification with Matrix Variate Normal and t Distributions

MixMatrix 0.2.6

• Update to fix M_PI issue with Rcpp update.
• Fix some non-HTTPS URLs.

MixMatrix 0.2.5

• Update with DOI for associated paper.
• Fix an issue with AR(1) and CS covariances. The way these are specified means that in case of mis-specification, they may be non-monotonic if they are mis-specified and results are near the the boundary. The functions were written such that it would not stop in this case.

MixMatrix 0.2.4

• Fixed a minor bug in DF calculation for mixture models.
• Fixed a minor bug in log likelihood calculation for mixture models.

MixMatrix 0.2.2

• Added more citations and references.
• Added reference to the paper this package accompanies.
• A small series of bug fixes and internal function updates to more easily support restrictions on means and variances different functions.

MixMatrix 0.2.1

• Added a number of citations and references to manuals and vignettes, including clarifying sources of any other functions.
• Added logLik functions for LDA, QDA, and MixMatrix objects

MixMatrix 0.2.0

• Start to add mixture modeling - currently support unconstrained variances only and fixed nu parameters for the model = "t" option..
• Remove list support for matrix LDA and QDA, since they were just a pain to support for the predict methods.

MixMatrix 0.1.0

• Preparing for CRAN submission
• Add pkgdown (https://gzt.github.io/MixMatrix/)
• Add ORCID
• Minor documentation changes
• Fix nu bug for matrixlda and matrixqda when using normal distribution
• Add vignette for t distribution

MixMatrix 0.0.0.9949

• changing name to MixMatrix

matrixdist 0.0.0.9927

• include EM algorithm (actually an ECME) for estimation of parameters of matrix-variate t-distributions
• include variance specifications for EM for t distribution
• include the possibility of restricting covariance matrices to a correlation structure.

matrixdist 0.0.0.9926

• split Wishart functions into CholWishart package
• migrate some internal functions to RcppArmadillo to improve speed
• add support for multiple observation input to dmatrixnorm function
• add support for multiple observation input to dmatrixt and dmatrixinvt
• dmatrix function now have large components in C++
• add error checking to dmatrixinvt - density only defined when matrix is positive definite
• improve speed of MLmatrixnorm by migrating more to C++ and fixing some R bottlenecks

matrixdist 0.0.0.9925

• add multivariate digamma
• add a couple tests for matrixt
• new plan: add fitting for t-distribution, LDA for t-distribution

matrixdist 0.0.0.9924

• Wrote density functions for the Wishart and InvWishart distributions. Broke the Wishart functions and mvgamma functions into a separate file just for them as they are all related and I may break them into their own package.

matrixdist 0.0.0.9923

• Rewrote rInvCholWishart to fix a bug where it was not actually a Cholesky decomposition - while upper triangular, tcrossprod() was InvWishart, not crossprod(). The method is a little slower since it involves computing rInvWishart and then taking the Cholesky decomposition, but it is faster than doing it in R.

matrixdist 0.0.0.9922

• Added LDA and QDA functions with predict() methods. These functions are relatively fragile.
• Marked some functions as internal to clean up the NAMESPACE

matrixdist 0.0.0.9921

• Added support for using IID structure for covariance matrices
• correction on restricted means - only true if known variance

matrixdist 0.0.0.9920

• added change to mean estimation to account for restricted means
• added changes to documentation

matrixdist 0.0.0.9919

• Added a NEWS.md file to track changes to the package.
• Changed the name of the maximum likelihood parameter fitting function to MLmatrixnorm since mle.--- could be confused for a method.

MixMatrix documentation built on Nov. 16, 2021, 9:25 a.m.