Re: Determination of sample size...again!

Dear Mr Wheeler,

Forgive my ignorance, but I do not understand what "hiding behind the
handle means". I would appreciate if you could be more explicit on
that matter.

Coming back to the experimental design, we expect that variabilities
in the two experimental conditions being tested will be quite similar
(sigma1 = sigma2). We are concerned about sample size simply because
we want to determine the number of donors to be collected for each
experimental condition that will allow us to detect a difference "D"
with a statistical power of, say, 80%, and alpha equal to 0.05.
Despite having consulted a couple of textbooks, I still cannot figure
out whether paired vs. unpaired studies require different sample size
calculations.

Jean-Francois Leblanc
Scientific Informatino Specialist
Hema-Quebec
Sainte-Foy, Quebec, CANADA (does this contact info solves the "hide
behind a handle" issue?)

Bob Wheeler <[EMAIL PROTECTED]> wrote in message news:<[EMAIL PROTECTED]>...
> As described it really isn't a sample size question, but rather one that 
> turns on the relative variablities.  But why in the heck should I help 
> you when you hide behind a handle?
> 
> Jean-Fran?ois Leblanc wrote:
> > Thanks to those who have provided answers to my everlasting questions.
> > 
> > To tell you a little bit more about our experimental design, we plan
> > to store whole blood in two different conditions, and evaluate the
> > effect of these two storage conditions on blood component quality. To
> > normalize for interindividual differences (which are often quite
> > great!), and because we need significant volumes of sample to assay
> > blood components as a function of time, we plan to collect two
> > compatible donations from two donors (one donation from each donor),
> > then immediately pool the two donations, and split them back into two
> > distinct blood bags to be stored in two distinct conditions. This
> > procedure has been reported in the literature for similar studies, and
> > in our case was chosen as an alternative to splitting a single
> > donation into two halves because, again, we need substantial volumes
> > of blood components for the different assays to be performed. This
> > experimental design is obviously paired. My question is: how this
> > pairing strategy affects sample size calculation? How should sample
> > size be calculated? What if an unpaired design (two separate
> > independent donations randomly assigned to either storage condition,
> > without "pool-and-split") would have been adopted? In such a case, how
> > is sample size determined?
> > 
> > Thanks again for your much appreciated help.
> > 
> > JF
> > 
> >