If you are going to make this program available for general
use you want to take every precaution to make it bulletproof.
This is a fairly informative data set. The model will undoubtedly
be used on far less informative data. While the model looks
pretty simple it is very challenging fr
010 9:48 AM
To: 'Rubén Roa'; 'Derek Ogle'; 'R'
Subject: Re: [R] optim() not finding optimal values
Ruben,
Transforming the parameters is also a good idea, but the obvious caveat is
that the transformation must be feasible. The log-transformation is only
feasibl
m: r-help-boun...@r-project.org [mailto:r-help-boun...@r-project.org] On
Behalf Of Rubén Roa
Sent: Monday, June 28, 2010 2:24 AM
To: Derek Ogle; R (r-help@R-project.org)
Subject: Re: [R] optim() not finding optimal values
Derek,
As a general strategy, and as an alternative to parscale when usin
mendi Ugartea z/g
48395 Sukarrieta (Bizkaia)
SPAIN
> -Mensaje original-
> De: r-help-boun...@r-project.org
> [mailto:r-help-boun...@r-project.org] En nombre de Derek Ogle
> Enviado el: sábado, 26 de junio de 2010 22:28
> Para: R (r-help@R-project.org)
> Asunto: [R] opti
11:52 PM
> To: Ravi Varadhan
> Cc: Derek Ogle; R (r-help@R-project.org)
> Subject: Re: [R] optim() not finding optimal values
>
> A slightly better scaling is the following:
>
> par.scale <- c(1.e06, 1.e06, 1.e-05, 1) # "q" is scaled differently
>
> >
unday, June 27, 2010 0:42 am
Subject: Re: [R] optim() not finding optimal values
To: Derek Ogle
Cc: "R (r-help@R-project.org)"
> Derek,
>
> The problem is that your function is poorly scaled. You can see
> that the parameters vary over 10 orders of magnitude (from 1e
@jhmi.edu
- Original Message -
From: Derek Ogle
Date: Saturday, June 26, 2010 4:28 pm
Subject: [R] optim() not finding optimal values
To: "R (r-help@R-project.org)"
> I am trying to use optim() to minimize a sum-of-squared deviations
> function based upon four parameters. The
Your function is very irregular, so the optim is likely to return
local minima rather than global minima.
Try different methods (SANN, CG, BFGS) and see if you get the result
you need. As with all numerical optimsation, I would check the
sensitivity of the results to starting values.
Ni
I am trying to use optim() to minimize a sum-of-squared deviations function
based upon four parameters. The basic function is defined as ...
SPsse <- function(par,B,CPE,SSE.only=TRUE) {
n <- length(B) # get number of years of data
B0 <- par["B0"]
9 matches
Mail list logo
