The geometric mean is the 'correct' statistic for log-transformed data (as opposed to the arithmetic mean). But it really depends on the distribution of the data. In the informal sense, if the population whose center you want to denote is symmetrical and fairly narrowly distributed (for example, the distribution of CD4 or CD8 on lymphocytes) then the geometric mean is not going to be too far away from the real answer. However, if the population is asymmetrical and/or broadly distributed (it likely contains several populations at too low a frequency to resolve well), then a better description of the center is probably the median (after you exclude the unlabeled cells). In the formal sense, if the log-transformed data are not normally distributed, then a parametric test such as a mean is nonsense. I wrote a paper on the this that you'll find on the dusty shelves of the library. Coder DM, Redelman D, Vogt RF Computing the central location of immunofluorescence distributions: logarithmic data transformations are not always appropriate. Cytometry. 1994 Jun 15;18(2):75-8. Dave dcoder@u.washington.edu ----- Original Message ----- From: Falco, Vincent <VFalco@Lifespan.org> To: cyto-inbox Sent: Monday, March 19, 2001 12:47 PM Subject: MEAN CHANNEL VALUE > > Group > > I have an investigator that wants to use mean channel value on a log scale > as a measure of changes in fluorescent intensity and thereby a surrogate > measure for changes in the surface marker antigen. Can I solicit any and all > thoughts and criticism. > Thanks in advance. > Vin Falco >
This archive was generated by hypermail 2b29 : Sun Jan 05 2003 - 19:01:12 EST