Implement derivation of expenditure share weights

[sweinand]

If prices and quantities are available, expenditure shares (weights) can be derived by region and/or product. This function may be used in cpd(), nlcpd() and prices().

shares <- function(p, q, r=NULL, n=NULL){

  # input checks:
  .check.num(x=p, int=c(0, Inf))
  .check.num(x=q, int=c(0, Inf))
  .check.char(x=r, miss.ok=TRUE, null.ok=TRUE)
  .check.char(x=n, miss.ok=TRUE, null.ok=TRUE)
  .check.lengths(x=p, y=q)
  .check.lengths(x=p, y=r)
  .check.lengths(x=p, y=n)

  # shares across all observations:
  if(is.null(r) && is.null(n)){
    s <- p*q/sum(p*q, na.rm=TRUE)
  }

  # shares by region:
  if(!is.null(r) && is.null(n)){
    s <- (p*q)/ave(x=p*q, r, FUN=function(z) sum(z, na.rm=TRUE))
  }

  # shares by product:
  if(is.null(r) && !is.null(n)){
    s <- (p*q)/ave(x=p*q, n, FUN=function(z) sum(z, na.rm=TRUE))
  }

  # shares by product and region, should result in a vector of ones:
  if(!is.null(r) && !is.null(n)){
    s <- (p*q)/ave(x=p*q, r, n, FUN=function(z) sum(z, na.rm=TRUE))
  }

  # return output:
  return(s)

}

[sweinand]

No implementation. Function is not needed as this can be easily solved by dt[, p*q / sum(p*q, na.rm=TRUE), by="region"] or dt[, p*q / sum(p*q, na.rm=TRUE)]