Re: data display

From: Dennis Broud 301-827-5246 FAX 301-594-6289 (BROUDD@cder.fda.gov)
Date: Tue Sep 30 1997 - 11:24:24 EST


What I find even more fun to look at are the 3-D plots, of course these also 
can be smoothed until they are as meaningless as any of the others.

Quite frankly, I think ye all doth protest to0 much.  The choice of how to 
present or display the data, depends on what you are trying to elucidate.  If 
one is looking for rare events, I've found the dot plot with the proper 
logical gates to be invaluable.  For separating subsets with very similar 
charactaristics, contour plots, with as little smoothing as necessary have 
been my choice.  With enough events, the 3-D plots can be very revealing 
particularly in combination with logical gating and 3 color staining.  The 
combination of contour and dot outliers (or is it outliars) sounds like fun, 
but I don't have that one available yet.  Histograms also have their place for 
certain kinds of data, energy transfer, FISH, etc.


Assuming that the original data is collected and compensated properly, all 
these data displays are merely tools allowing us to make interpretations of a 
mass of information.  Just as some data is better understood by expressing it 
on a log scale rather than a linear one, different types of displays allow us 
to demonstrate different points.  We all have our own little pet preferences 
for how we like to display the data.  While I like a good colorful multicolor 
presentation of data as well as the next flower, my administrative types, 
certainly do not like the page charges for color printing in journals and 
reprints.

I think that among the main questions are-- Does the data display support the 
interpretation that the author or speaker is expressing?.  Yes or no?  If it 
does, without stretching the bounds of believablity, then that's good enough.
There are times when one really does not want to get into all the little 
subgroups that a contour display might disclose.   

Not all of us chose to interpret data in the same manner, nor do we all have 
the same need and level of understanding.  Some people are splitters -- always 
looking for the differences, i.e. more and more subgroups, etc.  Other people 
are lumpers -- always looking for similarities and universality in data and 
across disciplines.  There is a room and a need for both.  If one has a real 
problem with a colleague's data, one can always repeat his experiments and 
subject it to his own analysis and interpretation, or be collegial and suggest 
that the data be analysed using a different display of the same data.  



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