data representation

Howard Shapiro hms at
Sat Aug 8 21:08:17 EST 1998

Jackie Saleh writes:

>Various users at my company are having very heated discussions on what
>number to use
>to beat reprent fluorescent intensity. Some users are using mean of M1(%
>positive cells)
>to describe fluorescent intensity. I believe this not is inaccurate because
>your using
>the mean of 20-30% of the cell population and I believe the mean of the
>entire population
>(not just M1) to be a much more accurate determination of fluorescence

If the sample includes positive and negative cells, and the user is
interested in the level of expression of an antigen on the positive cells,
then it makes more sense to consider a measure for just the positive cells
than a measure for the entire population.  The arithmetic mean is not
generally appropriate for distributions such as those typically encountered
in immunofluorescence measurement, which are skewed on a linear scale; the
geometric mean (which you get by doing an arithmetic mean on log data) is
somewhat better, but a statistician would probably prefer the median.

>I myself prefer to use percent positive cells as opposed to mean
>fluorescent intensity
>to describe my data.

That depends on the data; if you are lucky enough to be dealing with a
situation in which whatever experimental treatment you use makes cells
display an antigen they wouldn't have otherwise, the percentage of positive
cells gives the user directly relevant information.  If, on the other hand,
the treatment up- or downregulates an antigen which all or most of the cells
already have, you don't get very much information from the percentage of
positive cells.  

>I also explained to people that you cannot compare
>mean fluorescent
>intensities from this weeks experiment versus last weeks experiment because
>the means
>will vary considerably depending on were the negative control is set, if
>the planets are
>in line with Jupiter, and the "Flow Gods" see fit to smile on you.

If you don't calibrate and standardize your fluorescence measurements, you
are indeed at the mercy of the planets and the flow gods.  However, there
are several acceptable ways of calibrating immunofluorescence intensity
measurements, and an extended discussion of this topic will fill the October
1998 issue of Cytometry.


-- End --

More information about the Cytometry mailing list