TODO.md
In mlysy/LMN: Inference for Linear Models with Nuisance Parameters

[x] Uniform formatting of @title. Should be: first word only capitalized, ends with ".".
[x] @return lists. These should all be of the form

```r

' @return something something a list with elements: (don't forget the ":")

' \describe{

' \item{\code{a}}{Description of element a.}

' \item{\code{b}}{Description of element b.}

' }

```
[x] Latex code. Typically this doesn't display properly in vim/emacs & rstudio. So the trick is to define such commands twice, , e.g.:

```r

' \deqn{

' \boldsymbol{\Lambda}(\boldsymbol{Y}_t)^{\alpha - 1}

' }{

' \Lambda(Y_t)^(\alpha-1)

' }

```

To see the different outputs, compile the PDF doc (this can be done from the terminal using R CMD check LMN_XYZ.tar.gz), and also look at output in Rstudio (which is a bit nicer than vim/emacs, but still very basic and it's what most people will be using)
[x] @description. These can sometimes be made more informative, if the @title is not enough.
[x] Short examples for all exported functions (i.e., having the @export keyword). These go in a block called @examples. Also, the main doc for the package (in the file LMN.R) should have an example.
[x] Package vignette. A basic walkthrough, typically using a specific example. I'll get this started.
[x] Use documentation templates, i.e., man-roxygen and examples.
[x] Pass lmn_suff objects to other functions, instead of entire interface.
[x] Replace "." by "_".
[x] Convert roxygen documentation to Markdown formatting.
[x] Also, consider renaming Beta.hat to Bhat.
[x] Replace eqn and deqn roxygen tags with preformatted. Actually, no. Just write these properly. Not so hard.
[x] Remove debug arguments.
[x] ~~Package name -- should it be lmn???~~

No, I don't think this is worth doing at this point.
[x] Use solveV instead of solve for variance matrices.
[ ] Translate to C++ :)

Probably not worth the effort any time soon...
[x] Remove old files (under git version control anyways).
[ ] Remove src/DurbinLevison* methods, and call them from SuperGauss instead.

This means exposing the DL IP method in SuperGauss to call it from here...