G Robin Edwards <[EMAIL PROTECTED]> wrote: > In article <anetqm$[EMAIL PROTECTED]>, > Herman Rubin <[EMAIL PROTECTED]> wrote: > >> I would state otherwise; NEVER do any processing to your >> data not determined by the problem. Never normalize; >> never let the computer do your thinking for you. > > Thanks for this, Herman! > > It would be great if this opinion and advice could be communicated to > climatologists, who seem fixated on publishing their data in the form of > "normalised anomalies from the 1961 - 1990 average" - or some such > disguised version of the plain data, thus making things more difficult > for others (like me) who wish to examine their data further. Sometimes > they publish the "recipe" for the normalisation, but by no means always. > I can't understand what the advantage of "normalising" is supposed to be, > since any analysis will reach the same conclusions about relative > effects, normalised or not.
Usually in climatology the data set of interest has been laboriously assembled from many different partial data sets often measured at various distinct parts of the world at various times with widely varying experimental devices. There is usually some calibration necessary to ensure that long term changes observed or not come from actual geophysics instead of changes in the patterns of the technologies and techniques employed by various people over decades. . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
