MPN- method????

[Annette Nygaard Jensen]

Hi folks!

Please if anyone is into statistics - I need a piece of advice.

I have been estimating Most Probable Number, MPN due to a computerprogram 
by Hurley and Roscoe, 1983. When testing for significance between two MPN 
values they suggest a t-test using log MPN and standard error as in 
Cochran, 1950. However in Cochran the estimat for S.E. is given by a very 
simple equation which turn out to be a constant when dilution ratio and 
number of samples per dilution are constants - but the S.E. value 
provided from the computerprogram is not a constant - how come????
Furthermore there´s a lack of information about the distribution of the 
data etc. in order to argument for the use of a t-test, the calculation 
of 95% confidence intervals and the S.E. calculations.
Please if anyone can shed a little light on my statistic confusion I will 
be very thankfull. 

Annette N. Jensen
e-mail: anj at pop.bio.aau.dk
Institute of Microbial Ecology
University of Aarhus, Denmark 

Referred articles; 
M.A. Hurley & M.E. Roscoe. 1983. Automatedd Statistical analysis of 
microbial enumeration by dilution series. Journal of Applied Bacteriology 
vol 55, p159-164
W.G.Cochran. 1950. Estimation of Bacterial Densities by Means oif the 
Most Probable Number. Biometrics 6, p 105-116