[R-sig-Geo] [gstat] simple reboxing test fails

[Tom Roche]

https://stat.ethz.ch/pipermail/r-sig-geo/2012-November/016634.html
>> data3D.out <- krige(formula=v ~ 1, data3D.in, grid3D.out)
>> summary(data3D.out)
>> # gets
>> #> Object of class SpatialGridDataFrame
>> #> Coordinates:
>> #>      min max
>> #> [1,]   0   8
>> #> [2,]   0   8
>> #> [3,]   0   8
>> #> Is projected: NA
>> #> proj4string : [NA]
>> #> Grid attributes:
>> #>   cellcentre.offset cellsize cells.dim
>> #> 1               0.5        1         8
>> #> 2               0.5        1         8
>> #> 3               4.0        8         1
>> #> Data attributes:
>> #>    var1.pred    var1.var
>> #>  Min.   :1   Min.   : NA
>> #>  1st Qu.:1   1st Qu.: NA
>> #>  Median :1   Median : NA
>> #>  Mean   :1   Mean   :NaN
>> #>  3rd Qu.:1   3rd Qu.: NA
>> #>  Max.   :1   Max.   : NA
>> #>              NA's   :64

>> What should I call to get the values for the output grid boxes?

https://stat.ethz.ch/pipermail/r-sig-geo/2012-November/016638.html
> Did you notice the predictions in var1.pred values are all 1? 

Doh! no. I guess I've gotta convince myself that methods designed for
point prediction will work for grid interpolation. (I'd appreciate sound
arguments for an isomorphism--this may be obvious to geostatisticians
but I'm skeptical.) Or quit working while keeping one ear on the US
elections (at least that worked out better :-)

Onward to a more complex test ...

thanks, Tom Roche <Tom_Roche at pobox.com>