Spatial Statistics Lab 1


1. Get the built-in data set parana in geoR. Do an exploratory analysis of the data, compute an empirical variogram, and test for significant spatial correlation.
Useful functions:
summary
plot
points
variog
variog.mc

2. Fit a family of variogram models to the data. You decide which family to use. Set up a prediction grid, krige the data, display the fit and its standard error.
Useful functions:
variofit
likfit
expand.grid
krige.control
krige.conv
image
contour

3. Compare the variograms  from the fit in question 2 to those obtained using Cressie's robust variogram estimator (estimator.type="modulus" in variog), and a weighted least squares fit (using wei="cressie" in variofit).

4. You may have noticed a large spatial trend in question 1. What influence, if any, does this trend to have on the variogram fit? Consider two trend models (linear in coordinates, or quadratic in coordinates), Use an ordinary linear model to fit and compute residuals and variograms in both cases. Which model do you think is most appropriate?
Compare the ordinary kriging surface from question 2 to one computed using universal kriging (see documentation for type.krige in the function krige.conv to figure out how to do this).
Useful function:
lm

Some other functions you may find useful for this lab:
variog4
variog.model.env
plot.variogram
lines.variogram
lines.variomodel.likGRF
lines.variomodel