Skip to content

hclust graphics - plotting many points

3 messages · Karin Lagesen, michael watson (IAH-C), Hans Ekbrand

#
Hello.

I have a distance matrix with lots of distances that I use hclust to
organise. I then plot the results using the plot method of hclust.

However, the plot itself takes around 20 mins to make due to there
being ~700 things in the matrix that I have distances for. I thus
would like to dump this to some graphics format which will let me
examine this further.

I tried dumping it to postscript:

postscript("myfile.ps", height = 50, pointsize=5)
plot(my_hc_object)
dev.off()

What happens is that since most of the items in the matrix have a
distance of zero to something everything just becomes a black smear on
the bottom where I cannot distinguish anything from anything else. I
thus tried increasing the heigth and/or width and also downscaling the
pointsize. None of these improved anything much. 

So, now I am wondering if any of you have any tips for how I can get
something like I get in the x11() window which I can also store and
potentially show other people.

Thanks!

Karin
#
I'd recommend outputting either as pdf or as a windows metafile 

-----Original Message-----
From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org]
On Behalf Of Karin Lagesen
Sent: 10 March 2008 09:54
To: r-help at r-project.org
Subject: [R] hclust graphics - plotting many points


Hello.

I have a distance matrix with lots of distances that I use hclust to
organise. I then plot the results using the plot method of hclust.

However, the plot itself takes around 20 mins to make due to there
being ~700 things in the matrix that I have distances for. I thus
would like to dump this to some graphics format which will let me
examine this further.

I tried dumping it to postscript:

postscript("myfile.ps", height = 50, pointsize=5)
plot(my_hc_object)
dev.off()

What happens is that since most of the items in the matrix have a
distance of zero to something everything just becomes a black smear on
the bottom where I cannot distinguish anything from anything else. I
thus tried increasing the heigth and/or width and also downscaling the
pointsize. None of these improved anything much. 

So, now I am wondering if any of you have any tips for how I can get
something like I get in the x11() window which I can also store and
potentially show other people.

Thanks!

Karin
#
On Mon, Mar 10, 2008 at 10:19:01AM -0000, michael watson (IAH-C) wrote:
Don't you have the problem of too small distances in the X11() window?

I've had similar problems with a graph in graphviz, where I found it
easier to get what I wanted using a png-driver instead of postscript
driver.

Png doesn't scale well, but it might be worth a try.

png(file="myfile.png", width=3000, height=2250)
plot(my_hc_object)
dev.off()