Posted by
jo on
Aug 02, 2011; 7:55am
URL: http://imagej.273.s1.nabble.com/Re-question-regarding-Colocalization-Threshold-in-Fiji-Coloc2-tp3683647p3683651.html
>>> Robert Baer <
[hidden email]> 1.8.2011 23:49 >>>
-- snip --
> images for example one “positive” image where colocalization is
> expected/”visible” and one “negative” image with no colocalized
signals.
> Then, Íve tested the results with the Colocalization test plugin and
all
> the p values were much lower than 0.
the coloc test plugin randomises the image many times then compares
pearsons
r for each with the original.
we expect there to be lower pearsons r values than the real image pair
when
there is true coloc,
but about the same if the apparent coloc is no better than random.
read the Costes 2004 paper to get a full explanation.
Can you send me the images for testing?
>
> Do you really mean negative or is this a mis-speak? The theoretical
range
> of p is 0 to 1.0
no, the range for Pearsons r is -1 to +1
Sorry for this late answer:there was a mis-speak:the r(rand) were lower
than 0.The p value is 1.
OK, so the OP meant the correlation coefficient, r, and not
probability,
p, as he stated "p-value". In that case, I would agree that it was an
innocent mis-peak on the OP's part and not an algorithm issue.
Clearly, the
range for the correlation coefficient CAN BE -1 to 1, but NOT the range
for
probability.
So the only problem I have is the bug in the coloc2 plugin..
Thanks
G