"nlmnib" Package + Hessian Output

nserdar <snes1982 <at> hotmail.com> writes:

I need a "Hessian" matrix in "nlmnib" package to discuss whether parameters
are significant or not. 

Please let me know how to obtain hessian matrix and how to evaluate the
significancy of parameters.

You can get a finite-difference approximation to the hessian
by using the hessian() function (surprisingly enough) in the
numDeriv package.

  You haven't given enough detail about your problem, but you may
want to consider using the mle2() function in the bbmle package, which
will allow you to use nlminb as your optimizer but provides a wrapper
that computes significance tests (against the null hypothesis that 
individual parameters are zero).

Sorry but I don't modified my function with "mle2" :( :( 

Can you give example how to obtain Hessian with  numDeriv ?

Serdar 

######################### Function
Linn=function(param){

 phi1=((param[1]^2/(1+param[1]^2)))
 phi2=((param[2]^2/(1+param[2]^2)))
 phi3=((param[3]^2/(1+param[3]^2)))
 phi4=((param[4]^2/(1+param[4]^2)))
 
 sigw1=sqrt(exp(param[5]))
 sigw2=sqrt(exp(param[6]))
 sigw3=sqrt(exp(param[7]))
 sigw4=sqrt(exp(param[8]))
 
 sigv=sqrt(exp(param[9]))
 
 Betam1=((param[10]*100)/(sqrt(1+param[10]^2)))
 Betam2=((param[11]*100)/(sqrt(1+param[11]^2)))
 Betam3=((param[12]*100)/(sqrt(1+param[12]^2)))
 Betam4=((param[13]*100)/(sqrt(1+param[13]^2)))

phi=diag(c(phi1,phi2,phi3,phi4),4,4)
betam=c(Betam1,Betam2,Betam3,Betam4)
sigw=diag(c(sigw1,sigw2,sigw3,sigw4),4,4)

a<-(1.001)
mu0=c(ols[1,1],ols[2,1],ols[3,1],ols[4,1])
sigma0=diag(c(a,a,a,a),4,4)
  
 kf=kfilter1(n,rt,rm,mu0,sigma0,phi,betam,sigw,sigv)
 
 return(kf$like)
 
 }

a<-(1.001)
init.par<-c(0.5,0.5,0.5,0.5,a,a,a,a,a,ols[1,1],ols[2,1],ols[3,1],ols[4,1])

###########################






--
View this message in context: http://r.789695.n4.nabble.com/nlmnib-Package-Hessian-Output-tp4645768p4645838.html
Sent from the R help mailing list archive at Nabble.com.