Graphics Resolution

From: Howard Shapiro (hms@shapirolab.com)
Date: Wed Sep 24 1997 - 08:12:59 EST


This is motivated in part by the discussion forwarded from Joe Trotter about
WinMDI.
We all like to put the prettiest possible pictures into our presentations
and publications. However, what most of our flow cytometers and software are
generating are data based on 256- to 1024-channel one-dimensional
distributions and 64-by-64 to 256-by-256 channel two-dimensional
distributions.  These data are inherently somewhat "grainy", and are
therefore faithfully represented by relatively low-resolution bitmaps. What
higher display or output device resolution gets you is the ability to have
much better looking text for your legends and scale marking than you would
get with, say, 8-by-14 bitmapped characters.
Higher resolution does make contour plots look smoother, but a lot of us
tend to forget that a contour plot is typically made by smoothing a 64-by-64
or 128-by-128 two-dimensional distribution, and, as a result, looks much
better than the real data. A lot of the people who listen to our
presentations and read our papers never knew that in the first place.
Since I believe in showing the real data, warts and all, I avoid using
contour plots, and display two-dimensional distributions using a binary log
color scale or gray scale in which the color or gray value changes for every
doubling of the number of events (e.g., yellow for 1 event, blue for 2-3,
magenta for 4-7, green for 8-15, etc.).  Most software packages let you do this.
Pretty pictures are nice, but not to the point at which you end up with "ars
longa, data brevis".

-Howard  



This archive was generated by hypermail 2b29 : Wed Apr 03 2002 - 11:50:09 EST