Hello Confocalists
 
We are becoming more and more interested in determining quantitatively the
relationships between different classes of objects in three dimensional
space, as determined by 3D reconstructions of either confocal or
conventional fluorescence microscopy images. I am familiar with many areas
of statistics and stereology, as well as the newer methods for assessing
the true sizes and distributions of objects (disector type methods etc).
However, it seems that none of these methods by themselves really let you
determine whether an observed distribution of objects in 3D space is
random, clumped, or significantly associated with another class of objects.
I think this problem has been tackled to some degree by ecologists who need
to determine similar relationships betweeen organisms in forests etc, and
is based on various forms of nearest neighbour and / or cluster analysis. I
am more than happy to delve into this literature more carefully, and, if
necessary, try to develop some more appropriate methods. But before I
start, I am just wondering if anyone out there in microscopy land knows of,
or has used, this sort of statistical approach?
 
I should emphasise that this question is independent of the problems
associated with the accuracy of detection, the validity of the
reconstructions etc - although obviously the end result depends on how well
you can do those steps and, more important, how well you can estimate the
errors associated with them...
 
Thanks in advance,
IAN
Professor Ian Gibbins
Department of Anatomy and Histology
Flinders University of South Australia
Phone:  +61-8-2045271
FAX:    +61-8-2770085
e-mail:  [log in to unmask]