The content of this page has not been vetted since shifting away from MediaWiki. If you’d like to help, check out the how to help guide!
DEPRECATION NOTICE: This page describes the Colocalization Test plugin, which is no longer supported or maintained. Please use Coloc 2 instead.
on GitHub, modified from MBF ImageJ
Performs one of a set of three statistical tests, comparing the Persons correlation coefficient of 2 colour channels in the real image data against a white noise image, or the same image data with one of the colour channels spatially shifted, repeated a number of times.
See the great documentation for this plugin at Colocalization Analysis#Colocalization_Test.
This tells you if the colocalization that you measure with other plugins, such as Colocalization Threshold is better than random chance. For a busy image with lots of signal in both channels and very little area with no signal, there will be lots of random overlap, and this is easy to confuse with real colocalization with some biological meaning. It gives a P vaule (not a p-value) where 1 means all the randomised images had worse correlation than the real images, as you would expect for a real colocalised signel. A P value less than 0.95 is lower than the usual 95% statistical confidence limit.
The methods implemented are White Noise Image Approximation of Costes’ image randomisation (most robust, use 100 iterations), van Steensel and Fay (image shift methods). These differ in how the randomised images are generated from the real image or from white noise. None of them implement the Costes image randomisation test as described in his paper. Coloc_2 does though.
See also the Colocalization Analysis tutorial.