[R] Constrained Optimization, a full example

[Hesen Peng]

Hello,

Since the optimization goal is quadratic and constraint is linear,
quadratic programming functions will help. There are many such
packages available, e.g. quadprog.

On Tue, Jun 16, 2009 at 5:54 PM, Stu @ AGS<stu at agstechnet.com> wrote:
> After a few days of work, I think I nearly have it.
>
> Unfortunately, theta is unchanged after I run this (as a script from a
> file).  I thought that theta would contain the fitted parameters.
>
>
>
> The goal here is to find the least squares fit according to the function
> defined as "rss" subject to the constraints defined as ui and ci.
>
> I defined ui and ci to (hopefully) force par2 and par3 into the range  0.0
> <= par2 <= 1.0 and 0.0 <= par3 <= 1.0.
>
> I am not at all sure that ui and ci are defined correctly.
>
>
>
> The call to constrOptim returns normally, but without a solution
> (apparently).
>
>
>
> Any suggestions appreciated.
>
> Thanks
>
> Stu
>
>
>
> # data
>
> y <-
> c(0.111111111111111,0.0921052631578947,0.0564516129032258,0.0608108108108108
> ,0.0128205128205128,0.0136078431372549);
>
> x1 <-
> c(0,0.978723404255319,0.087378640776699,0.549295774647887,0.0596026490066225
> ,0.61578947368421);
>
> x2 <- c(1,3,4,5,6,7);
>
> x3 <- c(3600,169200,185400,255600,271800,342000);
>
>
>
> observs <- data.frame(y, x1, x2, x3);
>
>
>
> # function definition
>
> rss <- function(par, y, x1, x2, x3) {
>
>  par1 <- par[1]
>
>  par2 <- par[2]
>
>  par3 <- par[3]
>
>  ressq <- (y - par1 * (x1 + 1) * x2^(-par2) * x3^par3)^2
>
>  sum(ressq)
>
>  }
>
>
>
> #call to optimizer
>
> opti <- constrOptim(c(0.5, 0.5, 0.1), rss, NULL, ui = rbind(c(0, 0), c(1,
> 0), c(0, 1)), ci = c(0, 1, 1));
>
>
>
>
>        [[alternative HTML version deleted]]
>
> ______________________________________________
> 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.
>



-- 
彭河森 Hesen Peng
http://hesen.peng.googlepages.com/