[MLUG] R is nice

[Mike Miller <EMAIL:PROTECTED>]

We've been using R to make some stat graphs and I just thought I'd say a 
few words.  R is GPL and a top stat package.  All the academic 
statisticians are using it these days.

We were having some trouble getting what we wanted out of a scatter plot. 
We were trying to write to a png image, but it was tricky to control the 
sizes of elements like text and points and the fonts looked awful -- like 
small fonts had been amplified so that blocky pixelation was highly 
visible.  Eventually we realized that we had to use a vector graphics 
approach and convert to raster graphics using an external program.  This 
worked like a charm.

R allows use of postscript or pdf output devices.  We tried both.  The 
problem with postscript was that the aspect ratio didn't seem to hold when 
the image was displayed -- we were using kghostview and it would always 
stretch it out to fit the page size it was using.  There might be a way 
around this problem, but we found it easier to use the pdf output instead. 
Saving the output as graph.pdf, I used this ghostscript command to make a 
.png file from the .pdf file:

gs -q -dBATCH -dNOPAUSE -dLastPage=1 -sDEVICE=png16m -r333.333 -sOutputFile=graph.png graph.pdf

The graph is considered to be 6-inches (square) so the use of -r333.333 
gives us 333.333 dots per inch and a total of 2000x2000 pixels in the 
output .png file.  It can also do bmp or jpg or others:

http://pages.cs.wisc.edu/~ghost/doc/cvs/Devices.htm

We wanted an image for someone to import into PowerPoint (I guess it's OK 
with .png files now though it didn't used to be).  I think most people I 
know are just importing encapsulated postscript into TeX documents.  You 
can probably control the aspect ratio pretty easily then.

Mike

_______________________________________________
members mailing list
EMAIL:PROTECTED
http://mlug.missouri.edu/mailman/listinfo/members