RE: rare events

From: Howard Shapiro (hms@shapirolab.com)
Date: Mon Dec 06 1999 - 18:42:28 EST


Hai Qi reported finding 0.04% positives in one sample of 1/5 million cells 
(I assume this is 200,000 cells) and 0.15% in another.  In general, one can 
use Poisson statistics in dealing with rare event analysis (or with counts 
of small numbers of pretty much anything).  In these cases, 0.01% of 
200,000 is 20 cells, so the first sample had 80 positive cells and the 
second had 300.  The Poisson standard deviation for a count of n cells is 
the square root of n, which is about 9 for the 80 cells and about 18 for 
the 300.  The two values are thus separated by several standard deviations, 
which is to say that there is a statistically significant difference 
between them.  However, the statistics provide no information as to the 
source of the difference.  Since the cells came from the same pot, one 
would suspect instrumental factors related to data collection and/or 
analysis, unless there is reason to believe that a process such as 
differential settling of the rare cell type would change the composition of 
a sample aliquot with time.  A mild degree of paranoia is probably an asset 
when dealing with rare event analysis.

-Howard



This archive was generated by hypermail 2b29 : Wed Apr 03 2002 - 11:54:18 EST