R help - Jun 2008 - Minimizing the negative log likelihood function...

Hey,

The following is a function I wrote which generates random variables from a
Kappa (2-parameter) distribution.

rkappa <- function(n,beta,alpha){
 if(alpha <= 0)
  stop("alpha must be greater than zero!")
 if(beta <= 0)
  stop("beta must be greater than zero!")
 Vec <- beta*exp((1/alpha)*(log(-(alpha/(-1 +
exp(alpha*log(runif(n,0,1))))))+ alpha*log(runif(n,0,1))))
 return(Vec)
}

Now I would like to estimate the parameters of such a distribution using the
Maximum likelihood method.
I know that I have to minimize the following negative log likelihood
function:

Neg.Log.Like <- function(beta,alpha,x){
 -(sum( log((alpha/beta)*(alpha + (x/beta)^alpha)^( -(alpha + 1)/alpha))))
}

I have tried several R's functions for optimization but the results I yield
are not correct.
Is there anybody who can help me?

Thanks!

Francois Aucoin

	[[alternative HTML version deleted]]

Fran?ois Aucoin:
> I have tried several R's functions for optimization but the results I
> yield are not correct.
> Is there anybody who can help me?
I couldn't get it to estimate the correct values either, so I guess either
your random number generator 'rkappa' is wrong, or your
'Neg.Log.Like'
function is wrong. Or possibly both.

-- 
Karl Ove Hufthammer