R/repmeandif.R

In ILSAstats: Statistics for International Large-Scale Assessments (ILSA)

#' Mean Difference of Independent Samples with Replicate Weights
#'
#'
#' Estimates the mean difference for a single variable with replicate weights.
#' For a detailed explanation on how the standard errors are estimated
#' see \code{\link{repse}}.
#'
#'
#' @param x a data frame produced by \code{\link{repmean}} for a single variable.
#'
#'
#' @return a data frame or a list.
#'
#' @example inst/examples/repmeandif_example.R
#' @export
#'


repmeandif <- function(x){

  returnis(isrep.mean,x)


  if(is.data.frame(x)){
    message("dfs and pvalues are experimental.")
    return(.repmeandif(x))
  }


  message("dfs and pvalues are experimental.")
  lapply(x,function(i){
    .repmeandif(i)
  })



}

.repmeandif <- function(x){

  returnis(isrep.meansingle,x)

  # returnis(isrepmean,x)

  # if(min(c('N','mean','se')%in%colnames(x))==0)
  #   stop('Invalid input for x.')
  #
  # if(!is.data.frame(x))
  #   stop('Invalid input for x.')
  #
  # if('variable'%in%colnames(x))
  #   if(lu(unlist(x$variable))>1)
  #     stop('Invalid input for x. Please use only one variable.')


  # zou <- x[!x[,1]%in%'ALL',]
  zou <- x
  zou2 <- x[-2,]




  # composite
  exc <- attr(x,'excluded')
  grs <- zou2[-1,]
  grs <- grs[!grs$group%in%exc,]
  grs <- grs[!is.na(grs$mean),]
  mult <- ((nrow(grs))**2-1)/(nrow(grs)**2)
  comp <- sqrt(x[2,'se']**2+mult*grs[,'se']**2)

  # non composite

  if('variable'%in%colnames(x)){
    grn <- grep('variable',colnames(x))-1
  }else{
    grn <- grep('N',colnames(x))-1
  }

  if(grn==0)
    stop('Invalid input for x. No groups found.',call. = FALSE)


  mdif <- round(sapply(zou$mean,function(x) x-unlist(zou$mean)),5)

  mser <- round(sapply(zou$se,function(x){
    sqrt(unlist(zou$se)**2+x**2)
  }),5)

  ddff <- sapply(zou$N,function(x){
    (unlist(zou$N)-1)+(x-1)
  })

  mdif <- cbind.data.frame(dif = c(mdif),
                           se = c(mser),
                           tvalue = c(mdif/mser),
                           df = c(ddff))




  if(grn==1){
    nam <- unlist(zou$group)
  }else{
    nam <- apply(zou[,names(zou)[1:grn]],1,function(x) paste0(x,collapse = '_'))
    nam <- gsub(' ','',nam)
  }

  nam <- cbind.data.frame(rep(nam,each=nrow(zou)),rep(nam,nrow(zou)))
  colnames(nam) <- c('group1','group2')




  pv <- sapply(1:nrow(mdif),function(x){
    2*(stats::pt(q = abs(mdif$tvalue[x]), df = mdif$df[x], lower.tail = FALSE))

  })

  mdif <- cbind.data.frame(nam,mdif,pvalue = round(pv,5))

  mdiftot <- mdif[mdif$group1!='Composite',]
  mdifcom <- mdif[mdif$group1=='Composite',]

  mcom <- mdifcom[mdifcom$group2%in%grs$group,]

  mcom$se <- comp
  mcom$tvalue <- mcom$dif/comp



  out <- rbind.data.frame(mcom,mdiftot)
  rownames(out) <- NULL
  class(out) <- c("repmeandif", class(out))
  out


}