7.3 Spatial Statistics & Pattern Analysis

7.3 Spatial Statistics & Pattern Analysis

Human beings are naturally wired to find patterns, even where none exist. When looking at a map of crime incidents or disease outbreaks, visual interpretation alone is insufficient and highly subjective. What one analyst perceives as a "cluster" of incidents, another might dismiss as random noise. The Spatial Statistics toolbox in ArcGIS Pro removes this subjectivity by applying rigorous statistical theory to geographic data.

Spatial statistics allow us to quantify patterns, measure geographic distributions, and objectively determine whether an observed pattern is statistically significant or merely the result of random chance. For the EAPA_2025 exam, you must understand the distinction between descriptive statistics, global pattern analysis, and local cluster mapping.

Measuring Geographic Distributions

These tools are the spatial equivalents of traditional descriptive statistics (mean, median, standard deviation). They summarize the central tendency and spatial dispersion of a dataset.

Mean Center and Central Feature

• Mean Center: This tool identifies the geographic center (or center of concentration) for a set of features. It calculates the average X coordinate and the average Y coordinate to create a new single point. The resulting point might not fall exactly on an existing feature. It is the spatial equivalent of an average.
• Central Feature: Unlike the Mean Center, this tool identifies the single existing feature that has the lowest total distance to all other features in the dataset. The output is always one of the original input features. It is the spatial equivalent of a median.

Both tools can be weighted by an attribute. For example, calculating the Mean Center of cities weighted by their population will pull the center point toward the more populous cities.

Standard Distance and Directional Distribution

• Standard Distance: This tool measures the degree to which features are concentrated or dispersed around their geometric mean center. It outputs a circle. The larger the circle's radius, the more dispersed the features are. It is the spatial equivalent of standard deviation.
• Directional Distribution (Standard Deviational Ellipse): While Standard Distance assumes uniform dispersion in all directions, Directional Distribution captures the directional trend of the data. It outputs an elliptical polygon. The orientation of the ellipse indicates the primary directional trend (e.g., incidents spreading along a highway corridor), while the length of the major and minor axes represents the dispersion along those dimensions.

Analyzing Global Patterns

Global pattern analysis tools evaluate the entire dataset simultaneously to answer a single question: Is the overall pattern of this data clustered, dispersed, or random?

Average Nearest Neighbor

This tool measures the distance between each feature centroid and its nearest neighbor's centroid. It then averages all these nearest neighbor distances. If the average distance is less than the average for a hypothetical random distribution, the distribution is considered clustered. If the average distance is greater, it is dispersed.

• Use Case: This tool only evaluates feature locations (geometry). It does not analyze attribute values. It is useful for understanding competition (e.g., are retail stores clustered or dispersed?).

Spatial Autocorrelation (Global Moran's I)

Unlike Nearest Neighbor, Spatial Autocorrelation evaluates both feature locations and feature values simultaneously. It answers the question: Are features with similar values clustered closer together than features with dissimilar values?

• Mechanism: It compares the attribute value of a feature with the values of its neighboring features across the entire study area.
• Output: It returns a single Moran's Index value ranging from -1 to +1. A value near +1 indicates strong clustering of similar values. A value near -1 indicates strong dispersion (a checkerboard pattern). A value near 0 indicates complete randomness.

Understanding Statistical Significance (p-values and z-scores)

Global statistics return a z-score and a p-value. These are critical for interpretation.

• p-value: The probability that the observed spatial pattern was created by a random process. A p-value of 0.01 means there is only a 1% probability that the pattern is random. A small p-value (typically < 0.05) indicates the pattern is statistically significant.
• z-score: A measure of standard deviations from the mean. High positive z-scores indicate intense clustering, while high negative z-scores indicate intense dispersion.

If a tool returns a p-value of 0.50, you cannot reject the null hypothesis. The pattern must be treated as random, regardless of what it looks like visually.

Mapping Clusters (Local Statistics)

Global statistics tell you if clustering exists across the whole dataset, but they do not tell you where the clusters are located. Local statistics analyze the data at the neighborhood level to map specific clusters and outliers.

Hot Spot Analysis (Getis-Ord Gi*)

This is perhaps the most widely used spatial statistic. The Getis-Ord Gi* tool identifies statistically significant spatial clusters of high values (hot spots) and low values (cold spots).

• Mechanism: To be a statistically significant hot spot, a feature must have a high value and be surrounded by other features with high values. The local sum of a feature and its neighbors is compared proportionally to the sum of all features. If the local sum is much higher than expected, and the difference is too large to be random, it is designated a hot spot.
• Output: The tool assigns a z-score, p-value, and confidence level bin (e.g., 90%, 95%, 99% confidence) to every single input feature. This output is automatically symbolized on the map, making it exceptionally easy to interpret.
• Use Case: Identifying neighborhood blocks with statistically high rates of crime, or identifying agricultural fields with significantly low crop yields.

Cluster and Outlier Analysis (Anselin Local Moran's I)

While Hot Spot Analysis finds clusters, it masks spatial outliers. The Anselin Local Moran's I tool identifies both clusters and outliers.

• Clusters: It identifies clusters of high values (High-High) and clusters of low values (Low-Low).
• Outliers: Crucially, it identifies features with a high value surrounded by low values (High-Low outlier) and features with a low value surrounded by high values (Low-High outlier).
• Use Case: Identifying a wealthy household situated in a low-income neighborhood (High-Low outlier), or identifying an underperforming retail store located in an area where all other stores are highly profitable (Low-High outlier).

By integrating Spatial Statistics into your workflows, you elevate GIS from simple map-making to rigorous scientific analysis, a critical distinction tested heavily on the ArcGIS Pro Associate exam.