[R] Optim

[Ravi Varadhan]

The problem seems to be the fact that your mixture components are not
well-separated for your data, i.e. your data does not suggest two
components.  The third parameter, mixture proportion, converges to 1 (or 0)
from any sensible starting value.  

Ravi.


----------------------------------------------------------------------------
-------

Ravi Varadhan, Ph.D.

Assistant Professor, The Center on Aging and Health

Division of Geriatric Medicine and Gerontology 

Johns Hopkins University

Ph: (410) 502-2619

Fax: (410) 614-9625

Email: rvaradhan at jhmi.edu

Webpage:  http://www.jhsph.edu/agingandhealth/People/Faculty/Varadhan.html

 

----------------------------------------------------------------------------
--------


-----Original Message-----
From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org] On
Behalf Of mouradster
Sent: Thursday, February 12, 2009 12:54 AM
To: r-help at r-project.org
Subject: [R] Optim


Dear R user I follow the steps defined in Modern applied statistics
page(453) to use optim. However, when I run the following code the
parameters seems way off and the third parameter(p3) stayed as the initial
value.
below is the code:
## data
da=c(418,401,416,360,411,425,537,379,484,388,486,380,394,363,405,383,392,363
,398,526)

### initial values
pars=c(392.25, 507.25,   0.80)
p1=pars[1];p2=pars[2];p3=pars[3]
 
### Objective function
mix.obj = function(p,x)
{
	e
=p[3]*(pnorm((x+1/2-p[1])/sqrt(p[1]))-pnorm((x-1/2-p[1])/sqrt(p[1])))
 + (1-p[3])*(pnorm((x-1/2-p[2])/sqrt(p[2]))-pnorm((x-3/2-p[2])/sqrt(p[2])))
	if (any(e<=0)) Inf
	else -sum(log(e))
} 

### Calculate the first derivatives of the objective function.
lmix2a = deriv(~
-log(p3*pnorm((x+1/2-p1)/sqrt(p1))-pnorm((x-1/2-p1)/sqrt(p1))
 + (1-p3)*pnorm((x-1/2-p2)/sqrt(p2))-pnorm((x-3/2-p2)/sqrt(p2))),
c("p3","p2","p1"), 
	   function(x,p1,p2,p3) NULL)

mix.gr = function(p,x)
{
	p3 = p[3]
	p2 = p[2]
	p1 = p[1]
	colSums(attr(lmix2a(x,p3,p2,p1),"gradient"))
}
### optim
(resultsBFGS_D=optim(pars,mix.obj,mix.gr,x=da,method="L-BFGS-B",lower=rep(0,
3),upper=c(Inf,Inf,1),hessian=T)$par)

Also, sometimes it gives me the following error:
Error in optim(pars, mix.obj, mix.gr, x = da, method = "L-BFGS-B", lower =
rep(0,  : 
  L-BFGS-B needs finite values of 'fn'

Thanks in advance for ur help.

--
View this message in context:
http://www.nabble.com/Optim-tp21970390p21970390.html
Sent from the R help mailing list archive at Nabble.com.

______________________________________________
R-help at r-project.org mailing list
https://stat.ethz.ch/mailman/listinfo/r-help
PLEASE do read the posting guide http://www.R-project.org/posting-guide.html
and provide commented, minimal, self-contained, reproducible code.