1 EFFICIENCY
  Syntax: EFFICIENCY id1[:id1n][&idb1] id2[:id2n][&idb2] id3[:id3n][&idb3]
   where: id1,id2   are the input histogram identifiers
          id3       is the output histogram identifier
          idb1,idb2 are the (optional) input secondary identifiers
     and  idb3      is the (optional) output secondary identifier

  Divides two histograms  (`id1, id2') to make a  third one using binomial
  errors.  To  specify  the  secondary   identifier,  precede it  by a `&'
  otherwise the default will be used. (Use the `SET IDB' command to change
  the default).

  To  calculate the  efficiency  for a  range of  histograms,  the primary
  identifiers  you give for  the input and  output  histograms must be the
  same, but you can specify different  secondary identifiers. For example,
  `EFF 300:400&1 300:400&2 300 : 400 & 10' will divide all histograms with
  primary  identifiers 300 to  400 and  secondary  identifiers 1, by those
  with secondary  identifiers 2, putting the  results into histograms with
  the same primary  identifiers and  secondary  identifier 10. If you give
  primary identifier 0, the operation  will be performed on all plots with
  the given secondary identifier.

  The error on the efficiency is calculated using form derived by Paul
  Avery for MULFIT. This form makes some assumptions about the prior
  distribution, but is better than the naive one derived from a
  binomial.  The form is:

  sigma = sqrt [(n+1) * (N-n+1)] / [(N+3) * (N+2)^2]

  If the efficiency is 0 or 1 then the unbiased estimate of `n/N' is
  added to the error. This gives a better estimate of the confidence
  level:

  sigma = sigma + 1/(N+2)

  You should only use the `EFFICIENCY' command if the errors on the
  number of entries in each bin are Poisson distributed and the
  numerator is a subset of the denominator. If this is not the case
  you should use the `DIVIDE' command. It assumes that the numerator
  and denominator are uncorrelated when calculating the error.

