(Go back one page, or back to the Introduction).

Therefore, a properly compensated measurement of FITC fluorescence requires
the measurement of the fluorescence in the FITC channel as well as a measurement
in the PE channel. In our ideal world, every cell could be exactly compensated,
since all of these fluorescence could be measured exactly; since the fractional
spillover is the same for every cell, we would be able to compute the exact
contribution of FITC into PE (and vice versa) and end up with "pure"
measurements.

Indeed, in the real world, the instrument can do quite an excellent job
of compensation. However, there is a relatively small error in the ability
of the FACS to measure the fluorescence from any particular cell. And note,
from the equations above, that the error in computing the true compensated
fluorescence the error in the primary measurement PLUS the error in the
other channel multiplied by the compensation percentage. For very small
compensations, only a small error is carried over; the larger the compensation,
the greater the increase in the error.

This increased error especially affects cells that are dim in the compensated
channel: primarily, because the process becomes one of trying to measure
a small value above a large background (from spillover). Here, even small
measurement errors in the determination of the fluorescence in the other
channel will be significant with respect to the true signal we are trying
to measure.

The fallout of this error is a broadening of the distribution of the compensated
fluorescence (see example below): the error in the final measurement of
a fluorescence is the sum of the errors in measuring both fluorescences.

In a world of error-free measurements, compensation would not increase
the "width" (c.v.) of a population. This is because the amount
of fluorescence in the PE channel that arises from the FITC CD3 could be
exactly determined and exactly corrected. The width in the PE dimension
of the compensated population would be the same as for the uncompensated
sample (left panels). However, in the real world, the FITC measurement and
PE measurements are made with some inaccuracy. Therefore, it is impossible
to exactly determine the amount of signal in the PE channel that arises
from the FITC CD3. Thus, the compensated sample will have a width in the
PE dimension that is equal to the original width *plus an amount related
to the error of the measurement: i.e., it gets wider (right panels). Remember
that a single compensated fluorescence value depends on (in this case) *two
measurements: thus, the error in the compensation value will be the sum
of the errors in *both measurements.
*

Go on.