[R-sig-Geo] Likfit: Lambda values (Box-Cox's function)

[Paulo Justiniano Ribeiro Jr]

For certain data the model may be hard to identify
and objective function can produce odd values.

try to divide and conquer!

lambda = 0.01 is so close to lambda=0 which defines the log-transformation

Have you tried this value?

I would not try to fitt all parameters (at least at first)

the transformation parameter is ussualy almot orthigonal to the others so 
you could try to find the transformation first (use MASS:::boxcox())
and then run likfit fixint the transformation parameters.

For kappa I would also try to fit some values, 0.5, 1, 1.5 .... 5
first and chack the corresponding likelihoods.
In practice you may not need the fine tune on this parameter.

Paulo Justiniano Ribeiro Jr
LEG (Laboratorio de Estatistica e Geoinformacao)
Universidade Federal do Parana
Caixa Postal 19.081
CEP 81.531-990
Curitiba, PR  -  Brasil
Tel: (+55) 41 3361 3573
VOIP: (+55) (41) (3361 3600) 1053 1066
Fax: (+55) 41 3361 3141
e-mail: paulojus AT  ufpr  br
http://www.leg.ufpr.br/~paulojus

On Wed, 25 Mar 2015, Jimmy Neutrón wrote:

> Dear comRades:I'm working with the variogram "mattern". Then, I tried to run this function many times to achieve the best parameters for my geodata function (XXSelg) by means of likfit, then I wrote:likfit(XXSelg, cov.model="matern", ini.cov.pars=c(4.0, 1.8), kappa=3.5, fix.kappa=FALSE, nugget=0.8, lambda=0.01, fix.lambda=FALSE, hessian=TRUE)however, many times I received the following answer:likfit: end of numerical maximisation.Error en optim(par = 1, fn = .negloglik.boxcox, method = "L-BFGS-B", lower = limits$lambda["lower"],  : L-BFGS-B needs finite values of 'fn'
> Does 'lower = limits$lambda["lower"]' mean that I need lowed lambda values to achieve the best parameter?. I am even trying with negative values.
> Thanks in advance for you help.
> 	[[alternative HTML version deleted]]
>
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at r-project.org
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo
>