Attention all Statistics Fanatics

["Thomas Gray & Associates, Inc." <tga@tgainc.com>]

Hello All-

I have found myself in a statistical conundrum from which I cannot see the
light day!  Heres the problem.

I have a Bicron uRem meter attached to an ADC that is capable of sampling
the current at just about any rate one desires (ie..500 to infinity for
practical purposes).  I am trying to establish an a priori detection limit.
Because the ouput is dose rate and not counts, the standard Curry formula
based on the Poison Distribution for MDA cannot be directly applied.  My
thought was to assume that the each sampling of the current is a random
event and the sample size is such that a Gaussian Distribution is
approximated and therefore a T-test could be performed comparing two
theoretical means.  According to Curry, the resultant of such an analysis
only establishes a critical level and not a detection limit.  Here is an
example I test.

sample size 1=500
u 1= 5
std 1= 1
Sample size 2 = 500
u 2= 6
std 1= 1

Here is the result:

T- value = 15
Probability = 99.99%

My direct question is can a confidence level be estalished with the above
information so I could then say that the two populations of data are
different at a such and such confidence level.  Or if I am on the wrong
track with my line of thought can someone lead me down the path to
statistical bliss?

Thanks ahead of time to any who reply.

Matt Harmon
************************************************************************
The RADSAFE Frequently Asked Questions list, archives and subscription
information can be accessed at http://www.ehs.uiuc.edu/~rad/radsafe.html