R/multivariate.R

In pezzcvd/daphneg2:

#' multivariate
#'
#' #' Internal intermediate function. Starts the preprocessing procedure for
#' multivariate analysis. It differenciates among three possibilities (PxP, ExE, PxE),
#' phenotype target and phenotype covariate; environmental target and environmental covariate and
#' phenotypical target and environmental covariate respectively.
#' The covariate vector is selected among two options as well, either specifying one covariate
#' parameter (that must be present in the list of env/phn parameters) or considering the best
#' subselection among all parameters (this last option not available for PxP mode because of
#' data sparsity problem).
#'
#' @param mv.input character{1}. Name of the parameter of interest.
#' @param mv.par character{1}. Name of the covariate parameter.
#' @param mv.pheno data.frame. Either phenotypical or environmental.
#' @param genotype data.frame. Table with SNP occurrences among ecotypes.
#' @param mv.div FLag. If TRUE indicates the need of having both phenotypical and
#' environmental datasets (PxE analysis). (default F)
#' @param mv.env_altern data.frame. Environmental, loaded in case of PxE analysis.
#' @param mv.loco boolean. Leave one chromosome out approach. (default FALSE).
#' @param mv.pw character{1}. Output path. (default home folder).
#'
#' @return NULL. It's called by the preprocessing prodedure and it calls either "oneCov" or "nCov"
#' it's part of the preprocessing procedure of GWAS files in case of multivariate analysis.
#' @noRd
#'
#'
multivariate = function(mv.input, mv.par, mv.pheno, genotype,  mv.div = F, mv.env_altern = NULL, mv.loco = F, mv.pw = normalizePath("~")) {
  # Input controls
  checkmate::assert_character(x = mv.input, any.missing = F, len = 1)
  checkmate::assert_character(x = mv.par, any.missing = F, len = 1)
  checkmate::assert_data_frame(x = mv.pheno)
  checkmate::assert_flag(x = mv.div)
  checkmate::assert_data_frame(x = mv.env_altern, null.ok = T)
  checkmate::assert_character(x = mv.pw, any.missing = F, len = 1)

  # Filtering phenotype
  # Retrieves the data corresponding to the correct parameter and removes NAs
  input.line = which(colnames(mv.pheno) %in% mv.input )
  mv.x = mv.pheno[,input.line]
  mv.xn = which(!is.na(mv.x))

  # PxE case needs a second table for covariates
  # if pxe define cov table
  if (mv.div) {
    mv.table = mv.env_altern
  } else {
    mv.table = mv.pheno
    # Remove param of interest from the table that will be used for covariates
    parn = which(colnames(mv.pheno) == mv.input)
    mv.table = mv.pheno[,-parn]
  }

  # check case one covariate
  if (mv.par != "all") {
    oneCov(mv.input, mv.par, mv.table, mv.x, mv.xn, genotype, mv.loco, mv.pw)

    #check case more covariates --> linear model
  } else {
    nCov(mv.input, mv.table, mv.x, mv.xn, genotype, mv.loco, mv.pw)
  }

  return()
}