---
title: "Variograms, anisotropy, and external drift"
description: "Inspect spatial dependence, directional structure, trend assumptions, and covariate-supported kriging models."
output: rmarkdown::html_vignette
vignette: >
  %\VignetteIndexEntry{Variograms, anisotropy, and external drift}
  %\VignetteEngine{knitr::rmarkdown}
  %\VignetteEncoding{UTF-8}
---

```{r setup}
library(potentiomap)
data("synthetic_anisotropic_points")
p <- ps_make_points(synthetic_anisotropic_points, "x", "y", "head",
                    "point_id", "EPSG:26916")
```

```{r variogram}
v <- suppressWarnings(ps_variogram(p, directions = c(0, 45, 90, 135)))
head(v$empirical[, c("np", "dist", "gamma", "dir.hor")])
models <- suppressWarnings(ps_variogram_compare(v, c("Sph", "Exp")))
models$ranking[, c("model", "weighted_sse", "singular")]
a <- suppressWarnings(ps_anisotropy(p, minimum_pairs = 5))
a$summary
```

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.