[R] Gower Similarity Coefficient

[Martin Maechler]

>>>>> "Martin" == Martin Maechler <maechler at stat.math.ethz.ch>
>>>>>     on Mon, 14 Aug 2006 14:30:39 +0200 writes:

>>>>> "Chuck" == Chuck Cleland <ccleland at optonline.net>
>>>>>     on Mon, 14 Aug 2006 04:53:55 -0400 writes:

    Chuck> Timothy Rye wrote:
    >>> I'm interested in clustering my data using the Gower
    >>> Similarity Coefficient, and I was wondering if R is
    >>> capable of using that metric
    >>> 
    >>> Timothy Rye

    Chuck> RSiteSearch("Gower") points to a number of relevant
    Chuck> messages in the archives.  It also reveals gdist() in
    Chuck> the mvpart package, vegdist() in the vegan package,
    Chuck> and dist.binary() in the ade4 package, which may do
    Chuck> what you want.

    Martin> The daisy() function from the package "cluster"
    Martin> (which is *recommended* hence part of every complete
    Martin> R installation) is also based on Gower's
    Martin> (dis)similarity coefficient.

    Martin> So you don't need to install a new package [But I
    Martin> need to add the word 'Gower' to a better place on
    Martin> daisy()'s help page ...]

Hmm, actually, it's already there.

And if you use help.search() smartly 
[I use lib.loc = <R system library> in order to not search in
 the more than 1000 CRAN and bioconductor packages we have installed],

> help.search("Gower", agrep = FALSE, lib.loc = tail(.libPaths(), 1))

I get exactly the correct match

  >> Help files with alias or concept or title matching $,1rx(BGower$,1ry(B using
  >> regular expression matching:

  >> daisy(cluster)          Dissimilarity Matrix Calculation

  >> Type 'help(FOO, package = PKG)' to inspect entry 'FOO(PKG) TITLE'.


Martin Maechler