Rosetta - this is almost certainly due to the mathematics of rare-event analysis. The distribution of rare events follow Poisson statistics and the reproducibility of your estimates of "positive" events is determined by the number of positive events scored. You don't say how large your sample size was in each of the examples given but I would guess that the immunocytochemistry data are based upon a few thousand cells at best. Using Poisson statistics we know that the c.v. of the estimate of a reare event is given by the following formula - Cv = 100/ sqrt "n" (where "n" = the number of positive events E.g acquire a data file with 100,000 total events of which 100 are positive for your antigen of interest. This gives a frequency of 0.1% and the coefficient of variance of this result is 10%. Once you get below 100 positive events the c.v. increases exponentially and your precision decreases similarly. Dr Mark W Lowdell Senior Lecturer in Haematology / Head of Laboratory of Cellular Therapeutics RFUCMS T - 020 7830 2183 F - 020 7794 0645
This archive was generated by hypermail 2b29 : Wed Apr 03 2002 - 11:59:23 EST