thank you very much for your response!
But I have now another problem: optim( ) gives me the exactly same initial
values for the 4 parameters. I have tried many initial values, and it always
didnt change them.
opt$convergence = 1. I increased the number of iterations, but the problem is
still there. I also made sure that the function doesnt return "NAN" or
"INF" values...
Does anyone have any idea where the problem could come from ?
Many thanks.
> To: r-help@stat.math.ethz.ch
> From: bbolker@gmail.com
> Date: Tue, 12 Apr 2011 17:24:05 +0000
> Subject: Re: [R] question about optim
>
> chirine wolley <wolley.chirine <at> hotmail.com> writes:
>
> > Dear R-users, I would like to use optim( ) to minimize a function
> > which depends on 4 parameters: 2 vectors, a scalar, and a matrix.
> > And I have a hard to define the parameters at the beginning of the
> > function, and then to call optim. Indeed, all the examples I have
> > seen dont treat cases where parameters are not all real. Here is my
> > code, it doesnt work but its just to show you where is exactly my
> > problem:
>
>
> suppose the dimensions of X are (Xc,Xr)
> and the dimensions of x3 (I can't easily figure out what they should
> be) are (m,n)
>
> > g=function(x,matrice)
> > {
> > x1 = x[1:Xc] # 1st parameter x1 is a vector
> > x2 = x[Xc+1] # 2nd parameter x2 is real
> x3 <- matrix(x[(Xc+2):(Xc+1+m*n)],nrow=m,ncol=n)
> ## be careful about column- vs row-encoding, possibly use byrow=TRUE
>
> x4 <- x[(Xc+1+m*n+1):length(x)] #4th paramater is a vector
>
> > res1=rep(0,nrow(X))
> > res2=matrix(0,nrow=nrow(X),ncol=ncol(Y))
> > for (i in 1:nrow(X))
> > {
> > res1[i]=log(1/(1+exp(-t(x1)%*%X[i,]-x2)))
> > for (t in 1:ncol(Y))
> > {
> > res2[i,t]=log(((1+exp(-t(x3[,t])%*%X[i,]-x4[t]))/(sqrt(2*pi)))*
> > (exp(-0.5*((1+exp(-t(x3[,t])%*%X[i,]-x4[t]))*(1-Y[i,t]))^(2))))
> > }
> > }
> > sum(res1)+sum(res2) # the function to minimize
> > }
> >
>
> to call optim(), concatenate reasonable initial guesses for your
> parameters. (I have no idea what 'c' is in your code below ...)
>
> > opt=optim(c(alpha[,c+1],beta[c+1],w,gamma),g)
> > ### and how can we call optim ???
[[alternative HTML version deleted]]
