My problem is this: The Becton-Dickinson provides the D statistic and
the normalized D for use in hypothesis testing. The D lies between 0 and
1, always; the normalized D is a function of the sample sizes, and can
be any positive number. Dr. Kildare has been using the normalized D. I
explained that this was largely a measure of the number of B cells
entering the analysis, and not a measure of the difference between the
histograms. He should use the raw D, that lies between 0 and 1 (assuming
we are going to use this statistic at all.) Impossible, says Dr. Kildare:
the Mayo clinic uses the normalized D. "If they are, then they are mistake,"
I reply, "but I think they are using the raw D x 100. That is why their
"D" statistics are larger than 1." Dr. Kildare does not believe me; Mayo
clinic never mistaken, he claims; will not entertain possibility that they
are multiplying the D by hundred (which would be o.k.).
So can someone please tell me precisely what formula alleging to be that
of the Kolmogorov-Smirnov test statistic is being used by flow gurus at
the Mayo clinic, or a reference of Mayo provenance that I can show to
Dr. Kildare.
Thanks.
Catherine Dalzell,
beleagered statistician.