[R-sig-Geo] geoR questions

Paulo Justiniano Ribeiro Jr paulojus at c3sl.ufpr.br
Fri Jun 12 19:12:35 CEST 2009


I will try to answer some of your questions.
Ohters will depend on the intended resulkts and may not be suitable for 
discussions hers.

> Hello Everyone
> I am making my doctorate in biology at the University of Buenos Aires,
> Argentina.
> I am working on survival and dispersal of the south american fruit fly
> (Diptera: Tephritidae), a fruit pest mainly of citrics.
> To do so I have field dispersal data (with its x,y coordinates) and have
> already perform some spatial analysis with geoR.
> My specific questions are:
> 1- Data must show normality?  What to do if even transformed data does not
> adjust?

geoR (and "standard" geostatistical practices)
assumes data with are at least nearly normally distributed
(maybe after transformation)
Counts or binary data can be modelled by other models/alternatives.
An alternative for Poisson and Binomial data is implemented in geoRglm 

> 2- How do I interpret the cross-validation results?

cross-validation is **one of the methods** for judging or comparing models
based on comparing observed versus prediced values.
geostats literature is quite extensive on this topic

> 3- How do I know if I need to take into account anisotropy?

In my view anisotropy parameters are higly uncertain for most applications
and should be considered specially if you have good reasons to do so.
Adittionally you can compare fitted models with and without anisotropy to
assess the  tis relevance

> 4- When to end the iteration of likfit? till stable values of estimated
> parameters?

according to the convergance criteria in the optimization functions used
and the associated methids.
The default is optim() with L-BFGS-B but others can be set
see ?optim, ?nlminb

> 5- Is it recommended to present the image of the kriging AND the one
> representing variances? how do I generate the latter?

image() methods in geoR (as for persp and contour) have a "values"
argument which can point to the kriging variances.
See exemples in ?image.kriging and also in the tutorials
available at geoR page (www.leg.ufpr.br/geoR)

> 6- Do the kriging image values have a unit, or are they arbitrary?

dependes on the model but in general in geoR the same as you variable 
being modelled (this will depend if you have set a trabsformation 
parameter for the data)

> 7- Finally, if you had to present the results (in a paper for example)
> what results/data/graphics would you choose?

> Hoping to recieve lots(!) of answers,
> María Eugenia
> Lic. María Eugenia Utgés
> Bióloga
> Lab. Genética de Poblaciones Aplicada (GPA), Pab. 2, 4to piso, Lab. 105,
> Depto. Ecología, Genética y Evolución (EGE)
> Fac. Cs. Exactas y Naturales (FCEyN)
> Universidad de Buenos Aires (UBA)
> Ciudad Universitaria,
> Buenos Aires, Argentina
> Tel.: 4576-3300, int 218
> _______________________________________________
> R-sig-Geo mailing list
> R-sig-Geo at stat.math.ethz.ch
> https://stat.ethz.ch/mailman/listinfo/r-sig-geo

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
Fax: (+55) 41 3361 3141
e-mail: paulojus AT  ufpr  br

More information about the R-sig-Geo mailing list