[R-sig-Geo] How does variogramST handle NAs?

[Nick Hamm]

Dear Edzer

I was using the CRAN binaries (gstat version 1.0-15). The unexpected
result was that the variograms were totally different depending on
whether the NAs were included or not (the variogram decreased with
increasing temporal lag when NAs were excluded but increased when they
were included).

I installed gstat from R-forge (version 1.0-16).  I now get consistent
variograms.  I will look further at this next week.


Nick



On 3 February 2013 17:17, Edzer Pebesma <edzer.pebesma at uni-muenster.de> wrote:
> Dear Nick, did you use the r-forge binaries? This was suggested in
>
> https://stat.ethz.ch/pipermail/r-sig-geo/2013-January/017234.html
>
> Please let me know if your unexpected results persist, and also tell us what
> was unexpected.
>
>
>
>
> On 02/03/2013 01:40 PM, Nick Hamm wrote:
>>
>> Dear list
>>
>> We have been working with an STFDF object.  The dataset contains
>> several NAs, that is to say that there are some locations where the
>> time-series record is incomplete.  When we computed space-time
>> variograms for this dataset we found some unexpected results.
>>
>> To explore this further, we took a second dataset of measurements from
>> the same location.  This dataset is complete (i.e., there are no NAs).
>>   We computed the space-time variogram.  We then inserted NAs at the
>> location that they occur for the first dataset and get quite different
>> variograms!
>>
>> My question is:  How does variogramST handle NAs?
>>
>> Demo code is below.  You can download the data from here:
>>
>> https://dl.dropbox.com/u/15122401/demoData.zip
>>
>> Nick
>>
>>
>> sessionInfo()
>>
>>> sessionInfo()
>>
>> R version 2.15.2 (2012-10-26)
>> Platform: x86_64-w64-mingw32/x64 (64-bit)
>>
>> locale:
>> [1] LC_COLLATE=English_United Kingdom.1252  LC_CTYPE=English_United
>> Kingdom.1252
>> [3] LC_MONETARY=English_United Kingdom.1252 LC_NUMERIC=C
>> [5] LC_TIME=English_United Kingdom.1252
>>
>> attached base packages:
>> [1] stats     graphics  grDevices utils     datasets  methods   base
>>
>> other attached packages:
>> [1] gstat_1.0-15    spacetime_1.0-3 sp_1.0-5
>>
>> loaded via a namespace (and not attached):
>> [1] grid_2.15.2      intervals_0.13.3 lattice_0.20-10  tools_2.15.2
>> [5] xts_0.9-3        zoo_1.7-9
>>>
>>>
>>
>>
>> # Load required libraries
>> require(gstat)
>> require(spacetime)
>>
>> # Load the demo data
>> rm(list=ls())
>> load("demoData.RData")
>>
>> # jan.na has some NAs in time
>> summary(jan.na$value)
>> # LE is at the same location but has no NAs
>> summary(LE$value)
>>
>> # ST variogram for LE - full dataset
>> v.all <- variogramST(value~1, LE, width=20000, cutoff=500000,tlags=0:6)
>> plot(v.all, ylab="time lag (days)", xlab="distance (km)")
>> plot(v.all, map=FALSE, ylab="time lag (days)", xlab="distance (km)")
>>
>>
>> # Now insert NA's into the full dataset. The NA's correspond to  those
>> in the "jan.na" dataset
>> loc.na <- which(jan.na$value > 0)
>> LE.na  <- LE
>> LE.na$value[-loc.na] <- NA
>>
>> # ST variogram - dataset with NA's inserted
>> v.na <- variogramST(value~1, LE.na, width=20000, cutoff=500000,tlags=0:6)
>> plot(v.na, ylab="time lag (days)", xlab="distance (km)")
>> plot(v.na, map=FALSE, ylab="time lag (days)", xlab="distance (km)")
>>
>>
>>
>> # Remove all locations where there are any NAs in the jan.na time series
>> any.na <- which(apply(as(jan.na, "xts"), 2, function(x) all(!is.na(x))))
>> LE.no=LE[any.na,]
>>
>> # ST variogram - remove all locations where there is an NA
>> v.any <- variogramST(value~1, LE.no, width=20000, cutoff=500000,tlags=0:6)
>> plot(v.any, ylab="time lag (days)", xlab="distance (km)")
>> plot(v.any, map=FALSE, ylab="time lag (days)", xlab="distance (km)")
>>
>> _______________________________________________
>> 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