Re: Correlation coefficient of two channels for a gated population

Sydney Lu wrote:

>Thanks Dr. Shapiro. Given your explanation about the photoelectron 
>statistics, would it help matters at all to make sure that the two 
>parameters I wish to demonstrate a relationship for are detected 
>with antibodies coupled to fluors emitting at relatively short 
>wavelengths? I'm guessing PE-Cy7 and APC-Cy7 aren't the way to go here...

Yup.


>Additionally, if the situation (unstimulated) is a fairly tight 
>"cloud" that starts out uniformly spherical, r~0 and is situated at 
>say, 10^4 on both axes, and upon stimulus changes to a tight 
>45-degree slanted "rod" shape situated at around 10^4, does that 
>mean that there are relatively few worries?

Yes. If the intensities don't change much, there shouldn't be much 
contribution to variance from the photon statistics, and, if the 
shape of the cloud goes from near-circular to an ellipse with a very 
small minor axis, then you are on firm ground in claiming that the 
levels of expression are more highly correlated in the stimulated 
than the unstimulated cells. Figure 5-8 on p. 237 of the 4th Edition 
of Practical Flow Cytometry shows dot plots of highly correlated (r = 
.950) and uncorrelated (r = -.002) computer-generated data, should 
anyone reading this want help visualizing what is being discussed.

-Howard