
Re: log normal?
Posted:
Dec 31, 2012 6:12 PM


On Dec 27, Rich Ulrich <rich.ulr...@comcast.net> wrote:
> >> Ocasionally I come across somethign called log > >> normal, and I wonder, what is the purpose? > >> The normal distribution is natural, but the log of that, > >> seems unnatural, and unintuitive. > > >> Can anyone elaborate on its use? > > >Basically, the normal distribution is natural for additive > >processes. Add n uniformdeviate random numbers together > >m times, make a histogram and you get the Gaussian, i.e. > >normal, distribution. The log normal > >distribution is the equivalent for processes which are not > >additive but multiplicative. > > In detail with the algebra. > > Normal  > N comes from U1 + U2 + ... + Uk > > Log normal  > log(N) comes from log(U1) + log(U2) + ... + Uk > whichh implies N comes from U1 * U2 * ... *Uk
I didn't know this. But one thing is unclear  does that mean the final distribution is, mathematically, the log of a normal curve?
Like, "Class, on page 25, we see a normal Guassian curve, with specified mean and variance. Now let's compute the log of that, look on page 26, that's what log normal looks like."
Or does 'log normal' have some other arbitrary definition?
 Rich

