[R] Is there an R program that produces optimal solution/mix of multiple samples' varying volumes and values

[Mike Marchywka]

----------------------------------------
> Date: Mon, 25 Jul 2011 11:39:22 -0700
> From: lukescore770 at gmail.com
> To: r-help at r-project.org
> Subject: [R] Is there an R program that produces optimal solution/mix of multiple samples' varying volumes and values
>
> Sorry about the lengthy subject line.
> Anyone know of an R' program that can look at several sources' varying
> available volumes/amounts and their individual set of values, compared
> to a "target" range/curve for these values, to find the optimal
> mixture(s) of these possible sources for the desired curve, and for a
> specified amount? I hope that makes sense as a reader.

Well, whatever you are talking about you need to write some error as a function
of your fit parameters and decide if you can minimize it analytically or not.
If you can do it analytically, then there are probably matrix functions you want.
If not, there are several optimizers that should come up on google.

If you are trying to express some data in terms of nice basis set that
may be different than unknown collection of things. For example, if you
have sin/cos curves fft may work, polynomial something else etc. Often
when people ask questions like this, they try to fit to a collection of things
they think may work and then the optimizer gets stuck since it can't
optimize a and b for a*x + b*x etc so make sure your error function has a specific
"best" fit etc. Generally coding mistakes can be addressed on the list if you get that far.



>
> Thanks for your time.
> Luke