

   CClluusstteerr MMeemmbbeerrsshhiippss

        members(a)

   AArrgguummeennttss::

          a: output produced by functions `hierclust' (or
             'hclust' or `mclust').

   DDeessccrriippttiioonn::

        From a succession of agglomerations produced by a hier-
        archical routine in function `hierclust' or `hclust',
        determine cluster assignments of all objects, at all
        levels of the hierarchy.

   VVaalluuee::

        matrix of dimensions `n' by `n-2' giving cluster
        assignments to the 2, 3, ...{} n-1 cluster partitions
        of the hierarchy.  This corresponds to levels `n-2' to
        2.  The observations correspond to the rows, and are in
        sequence, `1, 2, ...{}'.  Column `j' specifies which of
        the `2, 3, ...{} j-1' clusters each observation is
        associated with at the level `j-1' of the agglomerative
        process.  The two clusters which merge in going from
        level `j-1' to `j-2' are replaced in the `j-2'nd column
        with the lower of the two cluster sequence numbers.

   NNOOTTEE::

        The time requirement of 'members' is `O(n^3)'.  Sample
        time for Sun SPARCstation 1, with 600 observations: 303
        secs.

   SSeeee AAllssoo::

        Functions `hierclust', `hclust', or `mclust' produce
        the array containing the sequence of agglomerations
        which can be used as input for `members'.

   EExxaammpplleess::

        h <- hierclust(a, method=2)
        k <- members(h)

