MathGroup Archive 1993

[Date Index] [Thread Index] [Author Index]

Search the Archive

lognormal distribution

  • To: mathgroup at yoda.physics.unc.edu
  • Subject: lognormal distribution
  • From: deb at alexandria.lcs.mit.edu (David E. Burmaster)
  • Date: Fri, 3 Dec 93 15:12:57 -0500

Dear MathGroup,

This message corrects several small errors in the message that I sent
only a few minutes ago.

Recently, I had to simulate several hundred realizations of a variable
distributed according to a lognormal distribution. I had parameterized
the problem such that 

        mu      =       average of the nat logs of the variable
        sigma   =       std dev of the nat logs of the variable

So, I found two ways to make a Table of realizations using one of 
Mma's standard packages....

In the first case, I used           

        LogNormalDistribution[mu, sigma]

and in the second case, I used

        Exp[NormalDistribution[mu, sigma]]

to do the realizations. Both give the same statistical results -- or so I think.

The first method take 50 to 55  times longer than the second method.

Am I missing something?? I think the two methods produce identical  
statistical results, but one runs 50 times faster. Do you agree?

Thank you for your help

Dave
David E. Burmaster 
Alceon Corporation
Cambridge, MA
USA






  • Next by Date: Bytes
  • Next by thread: Bytes