[R] Bivariate lognormal distribution

Vicky Landsman msvika at mscc.huji.ac.il
Fri Mar 25 18:13:00 CET 2005


Thanks to Prof. Ripley, Kjetil and Spencer Graves for help.
I will be more specific.
I have to simulate a bivariate lognormal pair (Y1,Y0) where E(Y1)=X'b, 
E(Y0)=X'd, Var(Y1)=c1, Var(Y0)=c0,
X is a data matrix, and b and d are vectors of parameters.
Vicky.


----- Original Message ----- 
From: "Spencer Graves" <spencer.graves at pdf.com>
To: "Prof Brian Ripley" <ripley at stats.ox.ac.uk>
Cc: "Vicky Landsman" <msvika at mscc.huji.ac.il>; "R-help list" 
<R-help at stat.math.ethz.ch>
Sent: Friday, March 25, 2005 4:40 PM
Subject: Re: [R] Bivariate lognormal distribution


>
>      I hope Professor Ripley will correct me if I'm mistaken, but the 
> documentation for "mvrnorm" in library(MASS) says it will, "Simulate from 
> a Multivariate Normal Distribution".  If you want the density function or 
> probabilities or quantiles, you can get those from library(mvtnorm).
>      Just for completeness, to use normal for a lognormal, you need to 
> take the logarithms of your number (which must be all positive;  zeros and 
> negative numbers become NA), then compute mean vector and variance matrix 
> of the logs, compute probabilities on the log scale, then back transform 
> by exponentiating to get the results back into the original scale.
>      hope this helps.  spencer graves
>
> Prof Brian Ripley wrote:
>
>> On Thu, 24 Mar 2005, Vicky Landsman wrote:
>>
>>> Is there a package that enables to create the bivariate log-normal 
>>> variables?
>>
>>
>> Just exponentiate each of a bivariate normal pair.  You can get the 
>> latter from mvrnorm in package MASS.
>>
>




More information about the R-help mailing list