Spatial Predictions Based on the Circular Variable-Radius Moving Window Method

A heuristic method to perform spatial predictions. The method consists of a local interpolator with stochastic features. It allows to build effective detailed maps and to estimate the spatial dependence without any assumptions on the spatial process.

spatialpred(coords, data, grid)

Arguments

coords	a data frame or numeric matrix containing columns with geographic coordinates
data	a numeric vector of compatible dimension with coords containing the response variable data to be predicted at each grid point
grid	a data frame or numeric matrix containing columns with geographic coordinates where data is to be predicted

Value

A data.frame containing spatial predictions, standard errors, the radius and the number of observations used in each prediction over the grid.

Details

If grid receives the same input as coords, spatialpred will calculate the Percenntual Absolute Mean Error (PAME) of predictions.

References

Da Silva, A.R., Silva, A.P.A., Tiago-Neto, L.J. (2020) A new local stochastic method for predicting data with spatial heterogeneity. ACTA SCIENTIARUM-AGRONOMY, 43:e49947.

Author

Anderson Rodrigo da Silva <anderson.agro@hotmail.com>

Warning

Depending on the dimension of coords and/or grid, spatialpred() can be time demanding.

See also

sHe

Examples

# data(moco)
# p <- spatialpred(coords = moco[, 1:2], data = rnorm(206), grid = moco[, 1:2]) 
# note: using coords as grid to calculate PAME
# head(p)
# lattice::levelplot(pred ~ Lat*Lon, data = p)

# End (not run)