Variograms, anisotropy, and external drift

library(potentiomap)
data("synthetic_anisotropic_points")
p <- ps_make_points(synthetic_anisotropic_points, "x", "y", "head",
                    "point_id", "EPSG:26916")

v <- suppressWarnings(ps_variogram(p, directions = c(0, 45, 90, 135)))
head(v$empirical[, c("np", "dist", "gamma", "dir.hor")])

##   np      dist        gamma dir.hor
## 1  2  29.47621 2.432955e-05       0
## 2  1 146.87500 1.444485e-02       0
## 3  3 221.59240 3.313863e-03       0
## 4  4 282.13430 4.181855e-02       0
## 5  6 388.00746 3.936191e-02       0
## 6  6 462.37606 8.419526e-02       0

models <- suppressWarnings(ps_variogram_compare(v, c("Sph", "Exp")))
models$ranking[, c("model", "weighted_sse", "singular")]

##   model weighted_sse singular
## 1   Sph           NA       NA
## 2   Exp           NA       NA

a <- suppressWarnings(ps_anisotropy(p, minimum_pairs = 5))
a$summary

##   major_continuity_direction minor_to_major_range_ratio usable_directions
## 1                         NA                         NA                 0
##                   angle_convention
## 1 clockwise from North; modulo 180

Directional evidence is exploratory and is not applied automatically. External drift must be supplied as scientifically defensible covariates available over the full prediction domain; association does not establish causality.