Dear Flowers, Please excuse my ignorance: I am presently having a little trouble understanding the statistics presented in flow software and have managed to get myself in a bit of muddle really so I would be grateful (not eternally so but certainly pretty grateful nonetheless) if someone could help me out.... Anyway straight to the point... WinList results windows display mean, medians and modal values. These are channel values (as I was keen enough to notice after having previouly used CellQuest) so to transform to log (expression) values and assuming I am using a 4 decade log scale at 1024 res I am performing the following calculation y=10^(x/256) where x is the average value you wish to transform Also I am assuming the mean quoted is the arithmetic mean which is equivalent to the geometric mean of transformed data. Is this assumption correct? Right well this seems fine but I am not sure where this leaves me with standard deviations. Are the standard deviations calculated directly from linear channel values or are they calculated from log data and the answer somehow converted into a linear channel value. I suppose my real question is how is the data stored in the list mode data file - I thought as linear channel values. If the standard deviation quoted by Winlist is calculated from the linear data (untransformed data) is this at all relevant in calculating error bars for expression (log) studies? I was hoping to use the STD as error bars around expression data but am not sure if this is 'correct'. I was establishing upper and lower limits on 1 standard deviation (by adding and subtracting STD from the mean expression), resulting in a linear value and transforming to find the log equivalents. Hoping someone can make some sense of my ramblings, Daniel Chipchase, Research Assistant, Lentivectors, Oxford Biomedica (UK) Ltd, Oxford Science Park.
This archive was generated by hypermail 2b29 : Sun Jan 05 2003 - 19:01:30 EST