Skip to content

why are idw and gstat different? [SEC=UNCLASSIFIED]

2 messages · Li Jin, Edzer Pebesma

#
Edzer,

Thanks for identifying the problem. The problem showed how big the impact of
the choice of nmax value on the estimation is.

In both krige and gstat, the default value for nmax is Inf. I am doing a
simulation experiment now to assess the performance of several spatial
interpolators using a few hundreds datasets. It is easy to do it by using the
default value. But I am wondering what the best guess for nmax is?

I assumed that the default value Inf for nmax would take the maximum value of
the number of samples available. After a few trials, I found my assumption
was wrong as evidenced below. Obviously, it took a value between 20 and 30.
[using ordinary kriging]
   user  system elapsed 
   0.32    0.00    0.31
[using ordinary kriging]
   user  system elapsed 
   0.25    0.02    0.27
[using ordinary kriging]
   user  system elapsed 
   0.41    0.02    0.43
[using ordinary kriging]
   user  system elapsed 
   0.89    0.00    0.89 

Please clarify this. 
Thanks,
Jin 

-----Original Message-----
From: Edzer Pebesma [mailto:edzer.pebesma at uni-muenster.de] 
Sent: Tuesday, 8 July 2008 5:35
To: Li Jin
Cc: r-sig-geo at stat.math.ethz.ch
Subject: Re: [R-sig-Geo] why are idw and gstat different? [SEC=UNCLASSIFIED]

Jin, you specified nmax = 7 in the second call, but not in the first. 
Comparing idp values is easiest when other things remain equal.
--
Edzer
Jin.Li at ga.gov.au wrote:
in
as