I am concerned that smoothing intoduces a false sense of security when analyzing merely uniparametric data. For example, when we first started measuring DNA contents in plants and obtained 3% CV's we thought this was really cool, and continued to accumulate for extended periods of time, maintaining the 3% CV. Incidentally, this indicates the staining is stable over time, but we just admired the way the jagged histograms became really smooth, which then were plotted manually by a graphic artist! Now smoothing allows presentations of histograms with alarmingly low ordinates, even with ordinates completely lacking range scales! Clearly the profiles appear smooth, but what do they signify? ...no sound or fury left, and, consequently, nada! David Galbraith P.S. 1,024 bin histograms only compounded the problem, since the numbers of events per bin were not compensated by an eight-fold increase in collection time!
This archive was generated by hypermail 2b29 : Wed Apr 03 2002 - 11:50:09 EST