Unbiased backtransformation of kriging predictions

Here you can suggest new features to be added to RGeostats package.

Unbiased backtransformation of kriging predictions

Postby Brian » Fri Mar 24, 2017 4:23 pm

Hi folks,

Thanks for developing this cracking R package, I've found it exceptionally useful for my research.

My work is focusing on the changing permeability of building masonry caused by stone decay. As the decay proceeds we see the development of areas of structural change leading to increased permeability. These high values mean that nearly all of my data needs to be transformed to remove asymmetry from the data.

I was wondering if functionality exists within RGeostats for unbiased back transformation after lognormal kriging? I've been using log10 to provide my distributions with a normal distribution prior to estimating the variograms.

#Transformation to log10 of value
data.db <- db.add(data.db,Slice.4.transformed=log10((Slice.4)+1))

Many Thanks

Brian Johnston
PhD Student
Queen's University Belfast.
Brian
 
Posts: 2
Joined: Thu Mar 09, 2017 10:48 am

Re: Unbiased backtransformation of kriging predictions

Postby Didier Renard » Sun Mar 26, 2017 3:17 pm

Hi Brian

Just to tell you that this \feature does not exist (... yet) but seems quite interesting.
As a matter of fact, the lognormal correction involves terms (such as the Lagrange coefficient) which is not present in the traditional kriging output. So including this correction in the future krigin procedure makes a lot of sense.

So I keep this post open just a reminder for next version's homework...

Thanks for your help
Didier Renard
 
Posts: 337
Joined: Thu Sep 20, 2012 4:22 pm

Re: Unbiased backtransformation of kriging predictions

Postby Brian » Mon Mar 27, 2017 2:11 pm

Cheers for the prompt reply. I'll look forward to future versions.

Many Thanks

Brian
Brian
 
Posts: 2
Joined: Thu Mar 09, 2017 10:48 am


Return to Global suggestions

Who is online

Users browsing this forum: No registered users and 0 guests

cron