Re: back-transformed values

[Edzer J. Pebesma]

Ulrich Leopold wrote:

Antonio Gallardo wrote:

   

Hi all,

I am starting to work with GSTAT under R, and I would like to know whether 
back-transformation (log transformation) of prediction results is possible 
under GSTAT. Any suggestion is welcome. 

Thanks in advance,



     

R can do the back-transform for you, if it's limited to simply calculating
the exponent. This gives you a median-type estimate on the original
scale. If you want an expectation-type estimate on the original
scale, you'd need the lagrange multiplier to do this and gstat does
not provide this as separate information; you'd have to modify it.
   

Is it difficult to modify the code?

I will need, to do POINT simulations of log-transformed data and
back-transform the points. Then aggregate the results to BLOCKS.
Ulrich

 

For back-transforming simulated values, you don't need any Lagrange
parameters; you only need it to back-transform the kriging estimate to
the log-normal kriging estimate.
--
Edzer