"Normalize.invariantset" in package "affy" is:

##  The 'common-to-all' part of the algorithm. Operates on two vectors of numeric data

##
normalize.invariantset <- function(data, ref, prd.td=c(0.003,0.007)) {

  np <- length(data)
  r.ref <- rank(ref)
  r.array <- rank(data)

  ## init
  prd.td.adj <- prd.td*10                           # adjusted threshold things
  i.set <- rep(TRUE, np)                            # index all the PM probes as being in the invariant set
  ns <- sum(i.set)                                  # number of probes in the invariant set
  ns.old <- ns+50+1                                 # number of probes previously in the invariant set

  ## iterate while the number of genes in the invariant set (ns) still varies...
  while ( (ns.old-ns) > 50 ) {
    air <- (r.ref[i.set] + r.array[i.set]) / (2*ns)  # average intensity rank for the probe intensities
    prd <- abs(r.ref[i.set] - r.array[i.set]) / ns
    threshold <- (prd.td.adj[2]-prd.td[1]) * air + prd.td.adj[1]
    i.set[i.set] <- (prd < threshold)

    ns.old <- ns
    ns <- sum(i.set)

    if (prd.td.adj[1] > prd.td[1])
      prd.td.adj <- prd.td.adj * 0.9  # update the adjusted threshold parameters
  }

Does anybody know that why should the prd.td be multiplied by 10 first?