R/OtimCov.R

In gcostaneto/envirotype: GIS platforms to support envirorytping in breeding programs

#'@title  Optimum number of covariates used as environmental markers
#'
#' @description Optimizes environmental variables matrix using PCA algorithms and Hierarchical Clustering
#' @author Germano Martins F Costa Neto <germano.cneto@usp.br>
#' @param .dfcov Kernel of environmental covariates (see envK function)
#' @param .size number of covariates per group
#' @param res.hcpc if TRUE perform the hierachical clustering
#' @importFrom FactoMineR PCA
#' @importFrom FactoMineR HCPC
#' @seealso envK
#' @seealso envKernel

OtimCov = function(.dfcov,.size=1,res.hcpc=TRUE){

  oticv<-c()
  Ws.red <- FactoMineR::PCA(t(.dfcov))
  Ws.red <- FactoMineR::HCPC(Ws.red,consol = T,graph = F)
  A<-Ws.red$call$X$clust
  for(i in 1:nlevels(A)){
    line<-sample(1:dim(Ws.red$call$X[Ws.red$call$X$clust == levels(A)[i],])[1], size = .size, replace = FALSE)
    oticv<-rbind(oticv,rownames(Ws.red$call$X[Ws.red$call$X$clust == levels(A)[i],])[line])
  }
  COV <- data.frame(cov=oticv)
  if(res.hcpc == TRUE){
    return(list(res.hcpc = Ws.red, newCov=.dfcov[,match(COV$cov,colnames(.dfcov))]))
  }
  if(res.hcpc == FALSE){
    return(.dfcov[,match(COV$cov,colnames(.dfcov))])
  }

}