Finding Hot Spots in ArcGIS Online: Minimizing the Subjectivity of Visual Analysis Nicholas M. Giner Esri Parrish S. Henderson FBI
Agenda The subjectivity of maps What is Hot Spot Analysis? Why do Hot Spot Analysis? Pattern Analysis: ArcGIS Desktop vs. ArcGIS Online Heat Maps vs. Hot Spot Maps How does Hot Spot Analysis work? How does Local Outlier Analysis work? Parrish Henderson Federal Bureau of Investigation
Maps are subjective Bronx, NY Water Quality Incidents January 2010 February 2017 Is there a spatial pattern in the location of these water quality incidents?
Maps are subjective Ft. Worth, TX Fire Response Times Is there a spatial pattern in the response times of these fire calls?
Maps are subjective Natural Breaks Southeast USA Economic Resilience to Natural Disasters (Counties) - 2010 Equal Interval Quantile Geometric Interval Is there a spatial pattern in economic resilience to natural disasters in the Southeastern USA?
What is Hot Spot Analysis? Identifies statistically significant hot spots and cold spots (i.e. spatial clusters of high and low values) in geographic data It is based on Tobler s First Law of Geography all things are related, but near things are more related than distant things (i.e. Spatial dependence or Spatial autocorrelation)
Why do Hot Spot Analysis? 1) Exploration: To reveal new insights and quantify patterns in data that you might not see 2) Regression workflows: Determining if you have spatial dependence in your residuals (one of the 6 checks of Ordinary Least Squares Regression ) 3) Interpolation workflows: Spatial dependence is the foundation of Geostatistics
Pattern Analysis: ArcGIS Desktop vs. ArcGIS Online ArcGIS Pro ArcGIS Online
Heat Maps vs. Hot Spot Maps
Demo #1 Heat Maps: Calculate Density in ArcGIS Online
How does Hot Spot Analysis work? Getis-Ord Gi* Statistic
How does Hot Spot Analysis work?
How does Hot Spot Analysis work? Feature
How does Hot Spot Analysis work? Feature Neighborhood
How does Hot Spot Analysis work? Study Area Feature Neighborhood
How does Hot Spot Analysis work? Study Area Is this Neighborhood Significantly different from the study area?
How does Hot Spot Analysis work? If significantly higher The feature is marked a Hot Spot Cold Spot 99% Confidence Cold Spot 95% Confidence Cold Spot 90% Confidence Not Significant Hot Spot 90% Confidence Hot Spot 95% Confidence Hot Spot 99% Confidence
How does Hot Spot Analysis work?
How does Hot Spot Analysis work?
How does Hot Spot Analysis work?
How does Hot Spot Analysis work?
How does Hot Spot Analysis work?
How does Hot Spot Analysis work? Cold Spot 99% Confidence Cold Spot 95% Confidence Cold Spot 90% Confidence Not Significant Hot Spot 90% Confidence Hot Spot 95% Confidence Hot Spot 99% Confidence
z-score (degree of clustering) How is Neighborhood defined? 6 5 Optimal neighborhood distance is where degree of clustering is highest 4 3 In ArcGIS Online, neighborhood is chosen for you via the Optimized Hot Spot Analysis tool Spatial Autocorrelation is calculated at multiple distances 2 1 20 40 60 80 100 120 140 Distance
Examples of Hot Spot Analysis Point Locations Chicago Crimes: (2014) DC Snow Removal Complaints: Jan 2016 February 2016 NYC Graffiti: Jan 2010- Present
Examples of Hot Spot Analysis Point Attributes Austria Heavy Metals: Cadmium Concentration
Examples of Hot Spot Analysis Polygon Attributes Philadelphia Tracts: Market Potential for Medicaid DC Block Groups: Republican Party Affiliation
Demo #2 Hot Spots: Find Hot Spots in ArcGIS Online
How does Local Outlier Analysis work? Local Indicators of Spatial Association (LISA) Statistic
How does Local Outlier Analysis work?
How does Local Outlier Analysis work? Feature
How does Local Outlier Analysis work? Feature Neighborhood
How does Local Outlier Analysis work? Is this Neighborhood Is this Feature AND Significantly different from all other neighborhoods? Significantly different from all other features?
How does Local Outlier Analysis work? Feature is higher than other features, Feature is higher than other features, Neighborhood is lower than other neighborhoods High Outlier HL HH Neighborhood is higher than other neighborhoods Feature is lower than other features, LL LH Feature is lower than other features, Neighborhood is lower than other neighborhoods Neighborhood is higher than other neighborhoods Low Outlier
How does Local Outlier Analysis work? High-High Cluster High-Low Outlier Low-High Outlier Low-Low Cluster Not Significant
Demo #3 Outliers: Find Outliers in ArcGIS Online
Analyzing Crime Hot Spots in New Orleans Parrish Henderson Federal Bureau of Investigation
Please Take Our Survey on the Esri Events App! Download the Esri Events app and find your event Select the session you attended Scroll down to find the survey Complete Answers and Select Submit
Print Your Certificate of Attendance Print stations located in the 140 Concourse Monday 12:30 PM 6:30 PM GIS Solutions Expo, Hall B Tuesday 10:45 AM 5:15 PM GIS Solutions Expo, Hall B 5:15 PM 6:30 PM Expo Social, Hall B 6:30 PM 9:30 PM Networking Reception, Smithsonian National Air and Space Museum
Appendix
Equations are local measures of spatial autocorrelation that calculate the degree of association between zone i and its neighbors j given a specified distance radius d (Getis and Ord, 1992). They help identify pockets of dependence called hot spots or cold spots. For example, if zone i has a high value and its neighbors j within distance d have high values, then this is a hot spot. The null hypothesis for these statistics is that there is no spatial clustering of similar values and the output is a map of z-scores. If a z-score is high or low (+ or 1.96 for a p-value < 0.05) then we can reject the null hypothesis and conclude that there is statistically significant spatial clustering of high or low values. where d represents the distance radius, the numerator is the sum of all zones j within d of zone i, and the denominator is the sum of all zones j in the study area not including zone i (Getis and Ord, 1992). The only difference between Gi and Gi* is that Gi* includes the value of zone i in the calculation (O Sullivan and Unwin, 2003). is a disaggregated version of the global Moran s statistic in that the equation is applied only to one particular zone rather than the summation of all the zones in the dataset (O Sullivan and Unwin, 2003). Thus it is a local measure of spatial autocorrelation and attempts to identify statistically significant clusters of similar values (high or low values). Because the LISA statistic is calculated locally, the Local Moran s index scores can be displayed spatially on a cluster map (Anselin, 1995). What is the definition of zone? In this case, the zone is defined by the adjacency matrix and the order of contiguity specified. For example, a firstorder, Queen s case contiguity zone around the state of Pennsylvania would include all states immediately sharing a border on all sides with Pennsylvania (Ohio, West Virginia, Maryland, Delaware, New Jersey, and New York). The zone would change if the type and order of contiguity changed. Other examples of the type of contiguity would be the Rook s and Bishop s case. Other examples of order of contiguity would be second or third order (Ord, 2010). where z i and z j represent deviations from the mean in zones i and j, and the summation over j indicates that only neighboring values are included in the calculation. W is the spatial weights matrix and defines adjacencies (Anselin, 1995).
z-scores and p-values