> I have a log normal distributin and i would like to know which is
> the best estimator of the mean. Is it the arithmetic mean (i don't think
> so), the median or the mean when the datas are transform in log?
If the distribution is given by density
f(x;m,s)=1/(sqrt(2*pi)*s*x) exp(-(log x-m)^2/2s^2), or log x~N(m,s^2),
then the expected value is exp(m+.5*s^2). That is what is estimated by the
sample mean. Median is exp(m), and it is a better thing to grasp an idea where
the distribution lies, especially if the variance of logs is large. m is also
the mean log. Without knowing what you are up to, I cannot say what you should
be using.
I have a one-page pdf file with the most useful properties of the lognormal
distribution compiled. I can make it available over the 'Net. There is also a
book by Aitchison and Brown dated back by 1950s.
......Stanislav Kolenikov (7-095)-232-3613p -3739f
[log in to unmask] http://giganda.komkon.org/~tacik/
..................Russian - European Center for Economic Policy
.................................................Moscow, Russia
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|