 |
 |
 |
 |
> Changing subject a bit, what bothers me is the occasional use of the cut
> line set on the control (e.g. 99% of control below the line) to estimate %
> positive when the entire distribution shifts slightly to the right (left
> edge as much as right edge) without changing shape. Although 20% may now
> be above the cut line, seems to me the result is that approximately 100% of
> the cells are (weakly) positive. The 20% is meaningless since the cells
> which were dimmest before staining have become just as much brighter as
> have the cells which were brightest before staining. Thus, seems to me the
> cut line should be used to estimate % positive as <100% only when the shape
> of the distribution changes, consistent with bimodality (at least a bump or
> tail on the right).
Eric,
I agree with you absolutely. I explain this to people as the Everest
hypothesis. i.e. Mount Everest is actually 2 mountains because half
is in Nepal and the other half in Tibet. We have used the shift in
median FI to describe these effects for several years. On the other
hand, when we have compared the median-shift method with just
counting the proportion of cells to the right of an arbitrary line
such as the 99th percentile of the control, the results for most
distributions show extremely close correlation. I think if people
use the "%+ve" value as an index in these cases, the data are quite
interpretable. If they think it actually means something at a
cellular level they are in trouble.
Have fun
Mike
...........................................................
Mike Salmon
Department of Rheumatology
The Medical School
University of Birmingham
Birmingham B15 2TT
United Kingdom
Tel: 44 (0)121 414 6780
Fax: 44 (0)121 414 6794
|
|
|
|