Dirichlet family

[Tanzmarie]

Dear gamboostLSS-Team,

I have implemented a novel Dirichlet family which allows for fitting Dirichlet regression models via gradient boosting within your gamboostLSS package for a paper that is currently under review. One referree pointed out that it would be favorable to add the family to gamboostLSS. To this end, I adjusted the necessary file in my fork and conducted relevant tests.

I would be really happy if you would consider adding the Dirichlet family.

Kind regards,
Michael

[hofnerb]

@mayrandy, it seems that our devel branch is completely outdated and that Michael (@Tanzmarie) has used that branch to develop his family. I am not sure if one can still change the merge target to master and check for differences there? I am currently having issues reviewing the code as it is based on many changes from the devel branch.

As a sidenote: We should consider what to do with the devel branch. It seems that we should either update it to the most current version or remove it completely to avoid issues. Maybe we can have a short call in the comming days?

[hofnerb]

@Tanzmarie: The code is currently not ok. Please run R checks and remove all notes, warnings and errors related to your code. A quick check gives me the following issues:

• DirichletAlpha: no visible binding for global variable 'loss'
  Undefined global functions or variables:
  loss

• checking Rd line widths ... NOTE
  Rd file 'families.Rd':
  \usage lines wider than 90 characters:
  DirichletLSS(K = NULL, a1=NULL, a2=NULL, a3=NULL, a4=NULL, a5=NULL, a6=NULL, a7=NULL, a8=NULL, a9=NULL,
  m1=NULL, m2=NULL, m3=NULL, m4=NULL, m5=NULL, m6=NULL, m7=NULL, m8=NULL, m9=NULL,
  s1=NULL, s2=NULL, s3=NULL, s4=NULL, s5=NULL, s6=NULL, s7=NULL, s8=NULL, s9=NULL,

  These lines will be truncated in the PDF manual.

• checking for code/documentation mismatches ... WARNING
  Codoc mismatches from Rd file 'families.Rd':
  DirichletLSS
  Code: function(K, a1 = NULL, a2 = NULL, a3 = NULL, a4 = NULL, a5 =
  NULL, a6 = NULL, a7 = NULL, a8 = NULL, a9 = NULL, m1 =
  NULL, m2 = NULL, m3 = NULL, m4 = NULL, m5 = NULL, m6 =
  NULL, m7 = NULL, m8 = NULL, m9 = NULL, s1 = NULL, s2 =
  NULL, s3 = NULL, s4 = NULL, s5 = NULL, s6 = NULL, s7 =
  NULL, s8 = NULL, s9 = NULL, ...)
  Docs: function(K = NULL, a1 = NULL, a2 = NULL, a3 = NULL, a4 = NULL,
  a5 = NULL, a6 = NULL, a7 = NULL, a8 = NULL, a9 = NULL,
  m1 = NULL, m2 = NULL, m3 = NULL, m4 = NULL, m5 = NULL,
  m6 = NULL, m7 = NULL, m8 = NULL, m9 = NULL, s1 = NULL,
  s2 = NULL, s3 = NULL, s4 = NULL, s5 = NULL, s6 = NULL,
  s7 = NULL, s8 = NULL, s9 = NULL, ...)
  Mismatches in argument default values:
  Name: 'K' Code: Docs: NULL

• checking Rd \usage sections ... WARNING
  Undocumented arguments in Rd file 'families.Rd'
  'K' 'a1' 'a2' 'a3' 'a4' 'a5' 'a6' 'a7' 'a8' 'a9' 'm1' 'm2' 'm3' 'm4'
  'm5' 'm6' 'm7' 'm8' 'm9' 's1' 's2' 's3' 's4' 's5' 's6' 's7' 's8' 's9'

  Functions with \usage entries need to have the appropriate \alias
  entries, and all their arguments documented.
  The \usage entries must correspond to syntactically valid R code.
  See chapter 'Writing R documentation files' in the 'Writing R
  Extensions' manual.

• checking for unstated dependencies in 'tests' ... WARNING
  'library' or 'require' call not declared from: 'DirichletReg'

• Running the tests in 'tests/regtest-families.R' failed.
  Last 13 lines of output:

  n <- 150
  p <- 10

  x <- matrix(runif(p * n, 0,1), n)

  x <- data.frame(x)

  a1 <- exp(2.5x[,1] - x[,2] + 3x[,3])
  a2 <- exp(2x[,4] + 2x[,5] - x[,6])
  a3 <- exp(1.5x[,7] - 1.5x[,8] + x[,9])
  A <- cbind(a1,a2,a3)

  y <- rdirichlet(nrow(A),A)
  Error in rdirichlet(nrow(A), A) : could not find function "rdirichlet"
  Execution halted

If possible you could also use the master branch and add your code to that to make the pull request easier as discussed above. I am also wondering if you really need the lengthy lists of arguments (a1, ...) which I currently don't even understand given a lack of documentation. Maybe we can see if it possible to use a vector or list of arguments instead?