The Important Methodologies
Point pattern analysis (event locations), uses include: : detect clustering/dispersion, hotspot detection, intensity modeling (Poisson/inhomogeneous Poisson)
Spatial autocorrelation and clustering, uses include: : quantify spatial dependence, find hot/cold spots, identify spatial outliers
Spatial regression (areal/point-referenced), uses include: model outcomes with spatial effects, mitigate biased inference, estimate spillovers
Geostatistics (continuous surfaces), uses include: : interpolate surfaces (pollution, rainfall), quantify uncertainty, optimal sampling
Distance/proximity and networks, uses include: neighborhood effects, street-network phenomena (crime, accessibility)
Spatiotemporal models, uses include: : track evolving clusters, forecasting, intervention evaluation
Change detection and surfaces, uses include: urban expansion, deforestation, regime shifts
Bayesian spatial modeling, uses include: small-area estimation, disease mapping, partial pooling with uncertainty
Spatial CV (Cross-Validation) is a crucial technique in spatial statistics and machine learning for evaluating model performance on geographic data, addressing issues like spatial autocorrelation by splitting data into spatial blocks (e.g., grids or clusters) instead of random partitions, ensuring that training and testing sets are truly independent and providing more realistic performance estimates, unlike random CV which often overestimates accuracy.
A hexbin (hexagonal binning) is a data visualization technique that groups 2D data points (like coordinates on a map or scatter plot) into hexagonal cells, coloring each hexagon based on the density or aggregate value of points within it, effectively showing data concentration and patterns by summarizing large datasets into a more readable, aggregated form, ideal for spotting trends in dense spatial or statistical data.
A proportional symbol map or proportional point symbol map is a type of thematic map that uses map symbols that vary in size to represent a quantitative variable.
Heatmaps using Kernel Density Estimation (KDE) are powerful visualization tools that transform clusters of point data (like crime incidents, website clicks, or scientific events) into continuous density surfaces, revealing hotspots and patterns by assigning intensity based on proximity and concentration, rather than showing individual points, offering confidentiality and clearer spatial trends.
Dot density (or dot distribution map) is a thematic mapping technique that uses point symbols (dots) to show the geographic distribution and density of a phenomenon, where each dot represents a fixed quantity (e.g., 100 people, 10 houses). It visualizes spatial patterns and concentrations without classifying data like choropleth maps, allowing viewers to see variations in density at a glance and even represent multiple variables with different colored dots.
Contour and isosurface plots visualize constant-value data in scientific and engineering simulations, with contours showing lines or colored regions on 2D planes (like elevation lines on a map) and isosurfaces showing 3D surfaces where a scalar quantity (e.g., temperature, pressure) is the same.
Choropleths are thematic maps that use varying colors or shades to represent statistical data across predefined geographic areas (like countries, states, or counties) to show spatial patterns, with darker or more intense colors often indicating higher values for things like population density, income, or election results.
Point pattern analysis (PPA) is a spatial statistics method used to describe, visualize, and model the spatial arrangement of points (like trees, crimes, or disease cases) on a map to understand the underlying processes
Contiguity, k-NN (k-Nearest Neighbors), and distance bands are spatial analysis concepts used to define neighborhood relationships for data, differing in how they select neighbors: Contiguity uses shared borders (e.g., Rook/Queen), k-NN selects a fixed number (k) of closest points, and distance bands define neighbors within a specific radius (distance threshold), often with weights decreasing with distance (like kernels) or using thresholds to create bands.
Outlier and influence diagnostics - In spatial statistics, outliers are observations that deviate significantly from their spatial neighborhood, while influential observations have a disproportionately large impact on model parameters and predictions.
Spatial autocorrelation measures how similar or dissimilar neighboring features in a dataset are, indicating if patterns are clustered (positive), dispersed (negative), or random (near zero), often using Moran's I index, which helps identify significant spatial patterns in geography, ecology, or economics.
A variogram is a fundamental geostatistical tool that quantifies the spatial variability and continuity of data, showing how the dissimilarity between data points changes with distance and direction (lag). It plots the average squared difference (semi variance) between data pairs against their separation distance, helping to understand if nearby points are more similar.
Trend surface analysis, also known as trend surface mapping, is a mathematical technique used in archaeology and environmental sciences such as geology and soil science. The method involves using low-order polynomials of spatial coordinates to estimate a regular grid of points from scattered observations such as archeological finds or soil survey results.
Common uses by domain:
- Public health: disease mapping, outbreak detection
- Environment: pollution interpolation, habitat suitability
- Urban/crime: hotspots, network-based clustering
- Retail/real estate: spatial hedonic pricing, market spillovers
- Agriculture: yield maps, variograms for sampling design, cokriging with soil covariates