[BioC] heatmap.2

[carol white]

Thanks Sean for your reply.

Actually the question was not on the distance measure. What I wanted to know how the clustering is performed on two dimensions. Is the dissimilary function (whatever it is, euclidean, pearson correlation, etc) is performed on all variables and then, on the observations or any other way?

Is it more clear?

Best,

--- On Wed, 4/28/10, Sean Davis <seandavi at gmail.com> wrote:

> From: Sean Davis <seandavi at gmail.com>
> Subject: Re: [BioC] heatmap.2
> To: "carol white" <wht_crl at yahoo.com>
> Cc: bioconductor at stat.math.ethz.ch
> Date: Wednesday, April 28, 2010, 3:55 AM
> On Wed, Apr 28, 2010 at 4:36 AM,
> carol white <wht_crl at yahoo.com>
> wrote:
> > Hi,
> > Does heatmap.2 function combine all variables into a
> single overall measure of dissimilarity between two
> observations as explained in The elements of statistical
> learning, Hastie et al, 2001, pp457? Does this function
> calculate the dissimilarity between observations and
> variables as follows?
> >
> >         N   N    p
> > 1/(N^2) sum sum   sum d(xij,xi'j)
> >        i=1 i'=1  j=1
> >
> > where N is the number of observations, p the number of
> variables, xi and xi' are two different observations, and d
> is the dissimilarity between two variables, respectively.
> >
> > Any relevant information is welcome.
> 
> Hi, Carol.
> 
> The first place to stop when asking these types of
> questions is the
> help system.  help(heatmap.2) shows that the default
> distance function
> used is "dist".  Checking help(dist) reveals that
> there are many
> options for distance measurement, but the default is
> "euclidean".
> There are a number of examples and even a couple of
> references.
> 
> Hope that helps.
> 
> Sean
>