Agriculture Reference
In-Depth Information
+n¼ length(samp)
+ index ¼ 1:N
+ incl ¼ rep(0,n)
+ for(i in 1:N) {
+ near < - which(diss[i,samp] ¼¼ min(diss[i,samp]))
+ incl[near] ¼ incl[near] + pi[i]
+}
+ var(incl)
+}
> ds < - as.matrix(dist(framepop[,2:3]))
> pik < - rep(n/N,N)
> spbalance(ds,pik,framesrs$id)
[1] 0.3290909
> framepop < - cbind(framepop,strataid2 ¼ floor(framepop$xc*10)*10+
+ floor(framepop$yc*10))
> kk < - (1/table(framepop$strataid2))[as.numeric(framepop$
+ $strataid2)+1]
> spbalance(ds,kk,str$id)
[1] 0.1111589
> spbalance(ds,pik,(1:1000)[srs_grts¼¼1])
[1] 0.1430303
Note that the spatial balance index can be also calculated with the command sb of
the package BalancedSampling. The reference manual can be downloaded from
http://cran.r-project.org/web/packages/BalancedSampling/BalancedSampling.pdf .
The results are, of course, consistent with the variance of the area of the poly-
gons, because our population is not concentrated in some zones, but is randomly
distributed across the entire study region.
The GRTS design produces a sample with specified
π k s, so the HT estimator
[or its continuous population analog (Cordy 1993 ; Stevens 1997 )] can be applied to
estimate the population characteristics.
Stevens ( 1997 ) provided exact expressions for the
π kl s in a particular case of
GRTS. However, these expressions unfortunately prevent the proper use of vari-
ance estimators based on the HT or Yates-Grundy-Sen estimators (see Eqs. 1.27
and 1.30 ) because they tend to be unstable if some
π kl s are very close to zero
(Stevens and Olsen 2004 ).
Stevens and Olsen ( 2003 ) proposed a contrast-based variance estimator for the
GRTS design that has some similarities to the smoothed estimator (Overton and
Stehman 1993 ). For an RTS design, Overton and Stehman ( 1993 ) considered a
smoothed contrast-based estimator of the form V SMO ¼ X k wd k y k y k
2 , where
wd k s are a set of weights and y k (i.e., the smoothed value for y k ) is assumed to be a
weighted mean of the values observed at a point and its nearest neighbors in the
tessellation. In Stevens and Olsen ( 2003 ), the single contrast y k y k
2 was
replaced with an average of several contrasts over a local neighborhood, which is
Search WWH ::




Custom Search