An embedded and charset-unspecified text was scrubbed... Name: not available Url: https://stat.ethz.ch/pipermail/bioconductor/attachments/20071013/ 421d3449/attachment.pl

John Lande wrote: > Dear BioC, > > I need to apply the function heatmap_plus to my matrices and I need to use > distance which are not euclidean, but something like pearson, or cosine. > It seems that the field distfun can be fitted with different metrics, but > none of my needs. do you know any other way to apply these distance in this > environment? > Hi, John. I'm not sure what you mean by "none of my needs" for distfun. Here is a simple example using cor(). > m=matrix(rnorm(1000),nc=20) > heatmap_plus(m) > heatmap_plus(m,distfun=function(x) {as.dist(1-cor(t(x)))}) Two points require attention. First, the distfun must return a distance-like object. as.dist() will often do the trick. Also, some functions will require a t() to get the correct results. Hope this helps. Sean

John Lande wrote: > thank you I will try out this, > > by this sentence "none of my needs" I mean that I didn't find cosine > correlation in the help pages for the function heatmap Ah. I see your point. Try: RSiteSearch('cosine correlation') The first and second hits mention the hopach package, which includes the function disscosangle(). This might be what you want, though I am not sure. Sean

Hi John The function heaplot in the made4 package will do this for you. We also added additional functionality to the function in the new release of made4. If you provide heatplot with a character vector representing grouping in your samples, it will add a colour bar beneath the dendrogram so you can easily see if your samples cluster. The function pretty.dend is an extension to this and will add multiple colour bars for each of your sample covariates. Regards Aedin > Date: Sat, 13 Oct 2007 14:50:04 +0200 > From: "John Lande" <john.lande77 at="" gmail.com=""> > Subject: [BioC] heatmap_plus and distances > To: bioconductor at stat.math.ethz.ch > Message-ID: > <c2ebc3880710130550q1bbe72e9gaf85bb55853ec82f at="" mail.gmail.com=""> > Content-Type: text/plain > > Dear BioC, > > I need to apply the function heatmap_plus to my matrices and I need to use > distance which are not euclidean, but something like pearson, or cosine. > It seems that the field distfun can be fitted with different metrics, but > none of my needs. do you know any other way to apply these distance in this > environment? > > > best regards > > > John > >