6 messages · Edzer Pebesma, Carlo Innocenti, Paulo Justiniano Ribeiro Jr
I would like to perform a gaussian conditional simulation on a variable that is highly skewed. Could someone suggest me how to transform my variable in a gaussian distribution? Thanks, Carlo
Carlo Innocenti, PhD ISPRA Istituto Superiore per la Protezione e la Ricerca Ambientale High Institute for Environmental Protection and Research Via Vitaliano Brancati, 60 00144 Roma Italy phone: +39 06 5007 4639 email: carlo.innocenti at isprambiente.it [[alternative HTML version deleted]]
one way is the rank transform: x = rgamma(100, 0.1) hist(x) y = qnorm((order(x)/(length(x) + 0.5))) qqnorm(y) but transforming your conditionally simulated values back will be tricky. And I'm not sure about the 0.5.
I would like to perform a gaussian conditional simulation on a variable that is highly skewed. Could someone suggest me how to transform my variable in a gaussian distribution? Thanks, Carlo
Edzer Pebesma Institute for Geoinformatics (ifgi), University of M?nster, Heisenbergstra?e 2, 48149 M?nster, Germany; +49 251 83 33081 Journal of Statistical Software: http://www.jstatsoft.org/ Computers & Geosciences: http://elsevier.com/locate/cageo/ Spatial Statistics Society http://www.spatialstatistics.info -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 490 bytes Desc: OpenPGP digital signature URL: <https://stat.ethz.ch/pipermail/r-sig-geo/attachments/20150428/a8c094a2/attachment.bin>
Thank you Edzer. It works, but I need the back transformation as well. What about the gaussian anamorphosis with hermite polynomials? I know that is a method used to normalize the data before to apply the conditional gaussian simulation. There is a package in R to transform and back transform with this method? Another method that is used in these cases should be the normal scores transformation: is it implemented in some package? Carlo 2015-04-28 22:32 GMT+02:00 Edzer Pebesma <edzer.pebesma at uni-muenster.de>:
one way is the rank transform: x = rgamma(100, 0.1) hist(x) y = qnorm((order(x)/(length(x) + 0.5))) qqnorm(y) but transforming your conditionally simulated values back will be tricky. And I'm not sure about the 0.5. On 04/28/2015 06:09 PM, Carlo Innocenti wrote:
I would like to perform a gaussian conditional simulation on a variable that is highly skewed. Could someone suggest me how to transform my variable in a gaussian distribution? Thanks, Carlo
-- Edzer Pebesma Institute for Geoinformatics (ifgi), University of M?nster, Heisenbergstra?e 2, 48149 M?nster, Germany; +49 251 83 33081 Journal of Statistical Software: http://www.jstatsoft.org/ Computers & Geosciences: http://elsevier.com/locate/cageo/ Spatial Statistics Society http://www.spatialstatistics.info
_______________________________________________ R-sig-Geo mailing list R-sig-Geo at r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Carlo Innocenti, PhD ISPRA Istituto Superiore per la Protezione e la Ricerca Ambientale High Institute for Environmental Protection and Research Via Vitaliano Brancati, 60 00144 Roma Italy phone: +39 06 5007 4639 email: carlo.innocenti at isprambiente.it [[alternative HTML version deleted]]
Thank you Edzer. It works, but I need the back transformation as well. What about the gaussian anamorphosis with hermite polynomials? I know that is a method used to normalize the data before to apply the conditional gaussian simulation.
I haven't seen this in R.
There is a package in R to transform and back transform with this method? Another method that is used in these cases should be the normal scores transformation: is it implemented in some package?
I think what I suggested below is commonly referred to as normal score transformation, maybe with the variation of taking out the 0.5 or having 1, there. GSLIB used to advocate this, and had a program for backtransforming, making certain assumptions about the tails. I'd recommend trying log or Box-Cox first, and then look into copulas.
Carlo
2015-04-28 22:32 GMT+02:00 Edzer Pebesma <edzer.pebesma at uni-muenster.de
<mailto:edzer.pebesma at uni-muenster.de>>:
one way is the rank transform:
x = rgamma(100, 0.1)
hist(x)
y = qnorm((order(x)/(length(x) + 0.5)))
qqnorm(y)
but transforming your conditionally simulated values back will be
tricky. And I'm not sure about the 0.5.
On 04/28/2015 06:09 PM, Carlo Innocenti wrote:
> I would like to perform a gaussian conditional simulation on a
variable
> that is highly skewed.
> Could someone suggest me how to transform my variable in a gaussian
> distribution?
>
> Thanks,
> Carlo
>
>
--
Edzer Pebesma
Institute for Geoinformatics (ifgi), University of M?nster,
Heisenbergstra?e 2, 48149 M?nster, Germany; +49 251 83 33081
<tel:%2B49%20251%2083%2033081>
Journal of Statistical Software: http://www.jstatsoft.org/
Computers & Geosciences: http://elsevier.com/locate/cageo/
Spatial Statistics Society http://www.spatialstatistics.info
_______________________________________________
R-sig-Geo mailing list
R-sig-Geo at r-project.org <mailto:R-sig-Geo at r-project.org>
https://stat.ethz.ch/mailman/listinfo/r-sig-geo
--
Carlo Innocenti, PhD
ISPRA
Istituto Superiore per la Protezione e la Ricerca Ambientale
High Institute for Environmental Protection and Research
Via Vitaliano Brancati, 60
00144 Roma
Italy
phone: +39 06 5007 4639
email: carlo.innocenti at isprambiente.it
<mailto:carlo.innocenti at isprambiente.it>
Edzer Pebesma Institute for Geoinformatics (ifgi), University of M?nster, Heisenbergstra?e 2, 48149 M?nster, Germany; +49 251 83 33081 Journal of Statistical Software: http://www.jstatsoft.org/ Computers & Geosciences: http://elsevier.com/locate/cageo/ Spatial Statistics Society http://www.spatialstatistics.info -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 490 bytes Desc: OpenPGP digital signature URL: <https://stat.ethz.ch/pipermail/r-sig-geo/attachments/20150429/b8d6f743/attachment.bin>
Perhaps you can also look at the Box-Cox family of transformation for which some cases such as log is a particular case You may look at MASS::boxcox() and/or boxcox related functions in geoR If log works well you benefit from the results of log-normal kriging.
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
Thank you Edzer. It works, but I need the back transformation as well. What about the gaussian anamorphosis with hermite polynomials? I know that is a method used to normalize the data before to apply the conditional gaussian simulation. There is a package in R to transform and back transform with this method? Another method that is used in these cases should be the normal scores transformation: is it implemented in some package? Carlo 2015-04-28 22:32 GMT+02:00 Edzer Pebesma <edzer.pebesma at uni-muenster.de>:
one way is the rank transform: x = rgamma(100, 0.1) hist(x) y = qnorm((order(x)/(length(x) + 0.5))) qqnorm(y) but transforming your conditionally simulated values back will be tricky. And I'm not sure about the 0.5. On 04/28/2015 06:09 PM, Carlo Innocenti wrote:
I would like to perform a gaussian conditional simulation on a variable that is highly skewed. Could someone suggest me how to transform my variable in a gaussian distribution? Thanks, Carlo
-- Edzer Pebesma Institute for Geoinformatics (ifgi), University of M?nster, Heisenbergstra?e 2, 48149 M?nster, Germany; +49 251 83 33081 Journal of Statistical Software: http://www.jstatsoft.org/ Computers & Geosciences: http://elsevier.com/locate/cageo/ Spatial Statistics Society http://www.spatialstatistics.info
_______________________________________________ R-sig-Geo mailing list R-sig-Geo at r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo
-- Carlo Innocenti, PhD ISPRA Istituto Superiore per la Protezione e la Ricerca Ambientale High Institute for Environmental Protection and Research Via Vitaliano Brancati, 60 00144 Roma Italy phone: +39 06 5007 4639 email: carlo.innocenti at isprambiente.it [[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
Thank you for your answers. In the meantime, I found the package RGeostats developed by Geosciences Research Center of MINES ParisTech on http://rgeostats.free.fr and at the moment I'm studying it. It seems promising with 3D geostatistics capability and function for gaussian anamorphosis transformation and back transformation. Carlo 2015-04-29 15:11 GMT+02:00 Paulo Justiniano Ribeiro Jr < paulojus at c3sl.ufpr.br>:
Perhaps you can also look at the Box-Cox family of transformation for
which some cases such as log is a particular case You may look at MASS::boxcox() and/or boxcox related functions in geoR If log works well you benefit from the results of log-normal kriging.
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, 29 Apr 2015, Carlo Innocenti wrote: Thank you Edzer.
It works, but I need the back transformation as well. What about the gaussian anamorphosis with hermite polynomials? I know that is a method used to normalize the data before to apply the conditional gaussian simulation. There is a package in R to transform and back transform with this method? Another method that is used in these cases should be the normal scores transformation: is it implemented in some package? Carlo 2015-04-28 22:32 GMT+02:00 Edzer Pebesma <edzer.pebesma at uni-muenster.de>: one way is the rank transform:
x = rgamma(100, 0.1) hist(x) y = qnorm((order(x)/(length(x) + 0.5))) qqnorm(y) but transforming your conditionally simulated values back will be tricky. And I'm not sure about the 0.5. On 04/28/2015 06:09 PM, Carlo Innocenti wrote:
I would like to perform a gaussian conditional simulation on a variable that is highly skewed. Could someone suggest me how to transform my variable in a gaussian distribution? Thanks, Carlo
-- Edzer Pebesma Institute for Geoinformatics (ifgi), University of M?nster, Heisenbergstra?e 2, 48149 M?nster, Germany; +49 251 83 33081 Journal of Statistical Software: http://www.jstatsoft.org/ Computers & Geosciences: http://elsevier.com/locate/cageo/ Spatial Statistics Society http://www.spatialstatistics.info
_______________________________________________ R-sig-Geo mailing list R-sig-Geo at r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo
--
Carlo Innocenti, PhD
ISPRA
Istituto Superiore per la Protezione e la Ricerca Ambientale
High Institute for Environmental Protection and Research
Via Vitaliano Brancati, 60
00144 Roma
Italy
phone: +39 06 5007 4639
email: carlo.innocenti at isprambiente.it
[[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
_______________________________________________ R-sig-Geo mailing list R-sig-Geo at r-project.org https://stat.ethz.ch/mailman/listinfo/r-sig-geo
Carlo Innocenti, PhD ISPRA Istituto Superiore per la Protezione e la Ricerca Ambientale High Institute for Environmental Protection and Research Via Vitaliano Brancati, 60 00144 Roma Italy phone: +39 06 5007 4639 email: carlo.innocenti at isprambiente.it [[alternative HTML version deleted]]