Statistics, scale and ratio? A case where K-S applies?

From: Christopher S Boyce (csboyce@beckman.com)
Date: Fri Oct 22 1999 - 17:07:46 EST


Hi All,

I am wondering what your thoughts are on this:

Suppose you  calculate  a ratio between two separate  FL distributions (the
median, mode, or mean) on the same cell line  (ratio 1).  The acquisition for
each cell line being set on the autofluorescent sample.   Now suppose you want
to compare ratio(1)  with the ratio(2)  of another cell line stained with the
same set of mabs (or multiple cell lines for that matter).  What are the
considerations here?

1)   How valid are the calculated  ratios given the type of scale used to
measure the range of FL?  Linear         conversions of log data?   Channel
number only?

2)   There are obviously some limitations here with comparing very different
cell lines?  Say  for example,  kidney   cell vs. liver cells, rather than
liver cell samples from multiple animals?

3)   Would a subtraction of the two distributions being compared, then a K-S
test comparing the subtractions of       two       different cell lines be an
alternative analysis?  Somehow this doesn't seem right.

4)   Is setting the instrument on the unstained sample for each line really
useful?  Why not just set on the highest      signal  of the whole bunch, and
let the chips fall for all cell lines (if of course all cells being tested  fall
within the     same analysis gate for scatter  signals)?  Different cell types
would make  this approach  difficult.

5)   And statistics?  What is the comparison we are making?  Ratio of two
distribution's medians compared to the   ratio of two distribution's medians---
ad nauseum for every cell line?   What might be the best statistical test to
approach the measure of significance?    How would variation fit into the
picture?

Any ideas, any comments would be greatly appreciated.

Thanks,

Christopher Boyce
Hybritech
San Deigo, CA



This archive was generated by hypermail 2b29 : Wed Apr 03 2002 - 11:54:09 EST