Nothing

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

- 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.

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

- 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.

- 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

- 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.

- 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

- changing name to MixMatrix

- 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.

- 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

- add multivariate
`digamma`

- add a couple tests for
`matrixt`

- new plan: add fitting for t-distribution, LDA for t-distribution

- 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.

- 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.

- Added LDA and QDA functions with
`predict()`

methods. These functions are relatively fragile. - Marked some functions as internal to clean up the NAMESPACE

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

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

- 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.

**Any scripts or data that you put into this service are public.**

Embedding an R snippet on your website

Add the following code to your website.

For more information on customizing the embed code, read Embedding Snippets.