Re: [R] Maximum Likelihood Estimation and Optimisation

[Spencer Graves]

It is not obvious to me what parameters in what model you want to fit. 
Function "optim" does very well with many different kinds of problems. 
If you just want to estimate parameters of a probability distribution, 
function "fitdistr" in library(MASS) will do that.  A couple of days 
ago, I needed to fit a "Pareto distribution of the first kind."  A 
search of "www.r-project.org" -> search -> "R site search" uncovered 
functions for a Pareto distribtion of a different kind.  So, I wrote the 
following and used them to check "fitdistr" and then to actually fit the 
distribution to data.

hope this helps.  spencer graves
#####################################################
dpareto <-
function(x, shape, x0, log=FALSE){
    dp <- if(log) (log(shape-1)-shape*log(x/x0)-log(x0))
    else ((shape-1)*((x0/x)^shape)/x0)
    dp[x<x0] <- 0
    dp
}
ppareto <-
function(q, shape, x0, lower.tail = TRUE, log.p=FALSE){
    q <- pmax(x0, q)
    if(log.p){
        if(lower.tail){
            return(log(ppareto(q, shape, x0)))
        }
        else return((shape-1)*log(x0/q))
    }
    else{
        S.q <- (x0/q)^(shape-1)
        if(lower.tail)return(1-S.q)
        else return(S.q)
    }
}
qpareto <-
function(p, shape, x0, lower.tail=TRUE){
    if(lower.tail) p <- (1-p)
    x0*exp(-log(p)/(shape-1))
}
rpareto <-
function(n, shape, x0)
qpareto(runif(n), shape, x0, lower.tail=FALSE)
fitdistr(rpareto(10000, 3, 1), dpareto, list(shape=2.5), x0=1)

########################################################
Harold Doran wrote:
Well, lm() produces an OLS solution, which are also MLE 
solutions for the fixed effects. I think this is an easy
way, although maybe not the best.
BHHH is a numerical approximation that can be used when 
a closed form solution is not available. It is less
sophisticated than Newton-Raphson.
Is this helpful?

 
------
Harold C. Doran
Director of Research and Evaluation
New American Schools
675 N. Washington Street, Suite 220
Alexandria, Virginia 22314
703.647.1628 

-----Original Message-----
From: Fohr, Marc [AM] [mailto:[EMAIL PROTECTED]
Sent: Thursday, July 10, 2003 10:17 AM
To: [EMAIL PROTECTED]
Subject: [R] Maximum Likelihood Estimation and Optimisation
Hello,

I want to calculate a maximum likelihood funktion in R in 
order to solve for the parameters of an estimator. Is there
an easy way to do this in R? How do I get the parameters and
the value of the maximum likelihood funktion.
More, I want to specify the algorithm of the optimisation 
above: BHHH (Berndt Hall Hall Hausman). Is this possible?
Thanks a lot for your help and best regards

Marc

-----------------------------------------------------------------------------
Marc Fohr, CFA
Equity Portfolio Manager
First Private Investment Management
Neue Mainzer Strasse 75
D-60311 Frankfurt/Main
Phone: ++49 - 69 - 2607 5424
Fax: ++49 - 69 - 2607 5440
Email: [EMAIL PROTECTED]
______________________________________________
[EMAIL PROTECTED] mailing list
https://www.stat.math.ethz.ch/mailman/listinfo/r-help
______________________________________________
[EMAIL PROTECTED] mailing list
https://www.stat.math.ethz.ch/mailman/listinfo/r-help
______________________________________________
[EMAIL PROTECTED] mailing list
https://www.stat.math.ethz.ch/mailman/listinfo/r-help