2013 Esri International User Conference July 8 12, 2013 San Diego, California Technical Workshop Spatial Pattern Analysis: Mapping Trends and Clusters Lauren Rosenshein Bennett Brett Rose
Presentation Outline DESCRIBE spatial patterns MINIMIZE subjectivity DISCOVER spatial clusters DEMOS Understanding Distributions Exploring patterns Crime in DC Multi-dimensional data analysis
What are Spatial Statistics in ArcGIS? A set of exploratory methods and techniques Specifically developed for use with geographic data - They incorporate space (area, length, proximity, orientation) directly into their mathematics They describe and model: - Spatial Distributions - Spatial Patterns - Spatial Processes - Spatial Relationships Spatial al statistics extend what the eyes and mind do intuitively to assess spatial al patterns, trends and relationships ps.
Why use spatial statistics? Spatial Statistics help us assess: Patterns Relationships Trends L&L Car Rental How we present our results (colors, class breaks, symbols ) can either enhance or obscure communication. L&L Car Rental
DESCRIBE spatial patterns
Extracting Information from Data
Making sense of our data
Describing Spatial Data Whenever we look at data, we immediately want to understand: - Mean - Standard Deviations - Whenever we look at spatial data, we think about the same things: - Centrality (mean) - Spatial distribution (standard deviations) - 8
Describing Spatial Data Questions - Which site is most accessible? - Where is the population center and how it is changing over time? - Which gang territories have the broadest spatial distribution, and do they overlap? - Is there a directional trend in the spatial distribution of social unrest? Is it changing over time? 9
Directional distribution Normal distribution 99% 95% 68% 1 = 68% of features Mean + 2 = 95% of features 3 = 99% of features
Demo Social Media Analysis
MINIMIZING subjectivity
Points on a map Crime in Lincoln, Nebraska How intense is the clustering? Where are the hot spots?
The subjectivity of visual pattern analysis Crime density surfaces Where are the hot spots? Where are the clusters?
Maps can lie Crime density surfaces Standard deviations Quantile Where are the hot spots? Where are the clusters?
Demo Minimizing Subjectivity: Hot Spot Analysis
Inferential statistics Are these crime just random acts? What is the chance that the patterns we see are the result of random processes? Is it just bad luck? Armed Conflict in Africa Crime in Lincoln, Nebraska
Inferential statistics The tools tell us how likely it is that the pattern is random - p-values - z-scores Armed Conflict in Africa Crime in Lincoln, Nebraska
What are p-values? What are z-scores? Probability theory p-values are probabilities z-scores are standard deviations How would you interpret a z-score of 2.58, with a p-value of 0.01? Given the z-score of 2.58, there is less than 1% likelihood that this cluster of high values could be the result of random chance z-scores can be mapped to specific p-values - p-value 0.01 = z-score +/- 2.58 - p-value 0.05 = z-score +/-1.96 - p-value 0.10 = z-score +/-1.65
DISCOVER spatial clusters
Mapping Spatial Clusters Mapping spatial clusters - Hot and cold spots - Spatial outliers - Similar features Where are the 911 call hot spots? Which countries face similar challenges? Which homes sold for more than expected?
Mapping Spatial Clusters Questions Tools Examples Where are the hot spots? Where are the cold spots? How intense is the clustering? Hot Spot Analysis Cluster and Outlier Analysis Where are the sharpest boundaries between affluence and poverty? Where are biological diversity and habitat quality highest? Where are the spatial outliers? How can resources be most effectively deployed? Cluster and Outlier Analysis Hot Spot Analysis Where do we find anomalous spending patterns in Los Angeles? Where do we see unexpectedly high rates of diabetes? Where are kitchen fires a higher-than-expected proportion of residential fires? Which locations are farthest from the problem? Which features are most alike? What does the spatial fabric of the data look like? Hot Spot Analysis Grouping Analysis Where should evacuation sites be located? Which crimes in the database are most similar to the one just committed? Which disease incidents are likely part of the same outbreak?
Cluster and Outlier Analysis of Voting Patterns Percentage of Obama and McCain votes.
Space-Time Cluster Analysis Defining the space-time window Create the space-time weights matrix file (.swm) Use the.swm to perform cluster analyses
Space-time vs Time Snapshots Time-slices can disconnect related features Space-time relationships are relative to each feature s date stamp
Space-Time Cluster Analysis Visualization Default symbology at 10.1 is a 2D display To see the temporal patterns: - Animate the results or - Use 3D symbols in ArcGlobe or ArcScene
Grouping Analysis Group 1: low income, low education, high unemployment, high crime Group 2: middle values for all variables Group 3: higher income and education, lower unemployment and crime Multi-dimensional data analysis simplifies complex data
Demo Grouping Analysis
Grouping Analysis Group HH Resilience Climate Governance Pop Density 1 Best Moderate Moderate Low 2 Moderate Worst Worst Moderate 3 Worst Moderate Best Moderate 4 Moderate Best Moderate Worst
Resources for learning more QUESTIONS? http://esriurl.com/spatialstats - Short videos - Articles and blogs - Online documentation - Supplementary model and script tools - Hot spot, Regression, and ModelBuilder tutorials - https://geodacenter.asu.edu/arc_pysal LBennett@Esri.com BRose@Esri.com