Re: Data Analysis: An old problem revisited

From: BIGOS@Darwin.Stanford.EDU
Date: Thu Feb 13 1997 - 05:15:13 EST


Several items to consider:

1) If you are trying to measure the change in the average population intensity, then using 
the ratio of the medians is a valid way to go.

2) Logamps aren't perfect, but the modern ones aren't "that" bad. We have done extensive 
measurements on our BD machines (Facscan, Facstar+, Vantage) and find that the error in 
their logamps is no greater than +-5%, except right at the lower and upper edges (the same 
is true for the MoFlo - we have not looked at a Coulter machine in years, but would be 
surprised if they are much different).  So in a ratio, the error shouldn't be greater than 
+-10%.

I am confused about your relative fluoescence scale. A four decade log amp should go from 
1-10K (or .1-1K, etc.), not 0-10K. This distortion of the low decade could influence the 
results.


3) In order to compare measurements from day-to-day on the same machine, standardization is 
important. We standardize by setting a "standard particle" (i.e. bead) to the same position 
on each channel for every experiment. This requires slightly different pmt voltages. The 
relative sensitivity = 1/gain, where gain = (V/Vo)**7.4, where Vo is the standard pmt 
voltage and V is the voltage used. This is an accurate measure of how well the instrument 
is aligned overall. A 20% difference can indicate setup problems.

4) It is important to know the inherent error in the biological system. Do several 
preparations of the same sample measured at the same time give the same results?

-Marty Bigos
Stanford Shared FACS Facility



This archive was generated by hypermail 2b29 : Wed Apr 03 2002 - 11:49:23 EST