Chronology Current Month Current Thread Current Date
[Year List] [Month List (current year)] [Date Index] [Thread Index] [Thread Prev] [Thread Next] [Date Prev] [Date Next]

Re: Semi-log and log graphs



Having read what others are saying about logarithmic graphs, I just wanted
to point out (and I'm sure that they know this) that it is incorrect to do
a linear fit on the transformed (i.e., "logarithimized") data, in any
case. It is incorrect even when the uncertainties are weighted, a.k.a.
weighted least-squares or equivalent.

I have seen that approach used, even by physicists, to get the decay
constant (or whatever) in the exponential equation.

The problem is, of course, that in the transformed domain the uncertainty
bars are skewed (the logs will extend the bars downward and shorten them
upward), and hence the data point does not lie atop a normal distribution
of uncertainty, and hence the "normal" linear fit methods do not work.

Also, there are not many tools that allow for uncertainty in both
dependent and independent variables.

Allen Brown