R help - Jun 2009 - Specifying ui and ci such that ui %*% theta - ci >= 0

Hi,

                I am a bit stuck on specifying ui and ci.  

I have read Lange's book ((1999) Numerical Analysis for Statisticians) to
his approach and unfortunately his descriptions were not helpful for me.

 

Here is what I have:

ui <- rbind(c(0, -1, 0), c(0, 0, -1))

ci <- c(0, -1, -1))

theta <- c(0.5, 0.5, 0.1)

 

                My goal is to feed these into constrOptim such that there is
no constraint on theta[1] and that theta[2] and theta[3] are both
constrained to fall between 0.0 and 1.0.

 

                Perhaps my values for ui and ci are incorrect for these
constraints.

 

                I have already solved the problem in another statistics
package so I know the actual solution to the coefficients.  But our ultimate
goal is to do additional work in R.  But so far, I have been unable  to
recreate our solution in R.

 

Thank you,

Stu

 

 


	[[alternative HTML version deleted]]

My understanding is that optim(method="L-BFGS-B") does box constraints
and constrOptim linear constraints.  You want box.

On 17/06/09 17:51, Stu @ AGS wrote:
> Hi,
> [...]
>
>
>                  My goal is to feed these into constrOptim such that there
is
> no constraint on theta[1] and that theta[2] and theta[3] are both
> constrained to fall between 0.0 and 1.0.
>
>
>
>                  Perhaps my values for ui and ci are incorrect for these
> constraints.
>
>    
	[[alternative HTML version deleted]]