TO DO
=====

* Add LICENSE file so that Github recognizes the license used (see
  Logolas for example).

  -- Added

* Add license info to README (see Logolas package for illustration).

  -- Added

* Add "Citing this work" section to README (see Logolas package for
  illustration).

  -- Added

* Merge Installation and Examples into a "Quick Start" section (see
  Logolas package for illustration).

  -- Done. Also added vignettes part at the end, similar to Logolas.

* Remove the unnecessary @importFrom statements; e.g., function
  CorShrinkData does not use "melt" from the reshape2 package anywhere
  in the code.

  -- Removed unnecessary statements

* Avoid using @import statements; use @importFrom instead. Ideally,
  you do not want any "import" statements in NAMESPACE, only
  "importFrom" statements.

  -- Converted all import to importFrom

* Move "test_unused" to "tests", and create a proper "testthat" suite
  of tests using the scripts you have in the test directory. See my
  varbvs package (https://github.com/pcarbo/varbvs) for an example of
  this.

* Remove the unneeded Suggests in the DESCRIPTION file (e.g., Biobase
  doesn't seem to be used anywhere.)

    -- Done. Biobase and a few other suggested packages are removed. Kept glasso and corpcor
     as they are competing packages to CorShrink.

* Consider moving some of the data sets in inst/exdata to "data" so
  that they can be loaded with the data() function instead of a more
  complicated call to system.file().

     --  Done. data/ folder created and inst/extdata folder is deprecated.

* Do you need the packrat file? Maybe it is a good time to get rid of
  it?
     -- The package broke when I removed packrat. Can packrat be put under .Rbuildignore?

* Suggestion: Merge CorShinkData and CorShrinkMatrix into one
  function, "CorShink". Then the first argument is "dat" that could
  either be a data frame or a correlation matrix; the former case, the
  sample correlation matrix is computed. In this way, the user only
  has to learn one function.

    -- This part is a bit complicated. What if the user has a symmetric data matrix. It may
       be unlikely, but would not want the function to read that as a correlation matrix.

* You know what I'm going to say about the plotting... return a
  ggplot2 object! :) See this very nice package:
  https://cran.r-project.org/package=ggcorrplot

   -- Done

* Also, it would be great to have a legend in the plot to know what
  the colours mean!

   -- Legend added

* Fix this message: "ash cor only and ash cor PD matrices are
  different". Is this an error? Warning? Also, difficult to read...

   -- This message has now been removed.

* Why does bootcorSE_calc return a data frame? Shouldn't it be a
  matrix?

   -- Now it should return a matrix.

* Fix use of unknown functions "tic", "toc" and "gradient" in function
  estimate_mixprop.

    -- The scripts containing these functions have been removed during trimming
       of the package.