Thoughts on Correlation
29.09.15 10:52

a) 3D problem: 
==============
we cannot resove Z, and within 0.8um confocal can't either.
Only qualitative analysis is possible.

b) Pearson:
===========
are the two signals changing proportionally?
(no thresholding necessary)

c) Manders:
==========
works with binary objects 
provides two output parameters:
	which area fraction of A overlaps with B
	which area fraction of B overlaps with A
Manders problem: in Coli, we don't have well-defined binary objects.
We would have to define two thresholds and would get 2 results.


d) Idea for qualitative comparison:
===================================
- Apply threshold so that a fixed portion (e.g. 30%) of the cell is 
  above threshold in either channel.
  Result is relative overlap of each channels's strongest 30% (0..1).
  
- Advantage: 
	- Threshold is easy to explain,
  	- we get only one overlap factor (as foreground area in
  both channels is equal: 0.3 x cell area)

- Options:
	Analyse entire cell, or cell wall only

Software Distribution:
======================
as separate macro. User is free to run it from a separate text window, 
or to glue it to the end of the embedded macros.