Area Hot Spot Analysis in ArcGIS Online

This tutorial is an example of hot spot analysis with areas in ArcGIS Online. An example of hot spot analysis with points is given in the tutorial Crime Hot Spot Analysis in ArcGIS Online.

Mapping Areas

This tutorial uses 2010-2014 cancer data by census tract from the New York State Department of Health. The video below shows how to create a choropleth of breast cancer rates using an existing layer in the FSC ArcGIS Online organization. That data is also available as a Geo JSON file or as HTML from the New York State Department of Health.

Note that rather than providing directly-mappable incidence rates, the attributes for the different types of cancer are actual counts and expected counts (which are based on age-adjusted incidence rates). This requires dividing the counts by the expected counts to get a ratio that indicates whether the area has a higher or lower incidence of cancer than would be expected if cancer occurred in this area at the same rates as the rest of the country.

Creating a Breast Cancer Choropleth

Hot Spot Analysis Tool

In viewing the map of Long Island created above, it is clear that much of the island is covered with red areas, indicating higher than normal breast cancer rates. However those red areas are interspersed with blue areas indicating lower than normal breast cancer rates. While it is possible to use a map to identify specific areas with higher rates, simply observing a map of complex data like this does not offer clear, rigorous insight into general patterns.

The Getis-Ord GI* statistic was developed by Arthur Getis and J.K. Ord and introduced in 1992. This algorithm locates areas with high values that are also surrounded by high values. However, unlike simple observation or techniques like kernel density analysis, this algorithm uses statistical comparisons of all areas to create p-values indicating how probable it is that clusters of high values in a specific areas could have occurred by chance. This rigorous use of statistics makes it possible to have some confidence (but not complete confidence) that the patterns are more than just a visual illusion and represent clustering that is worth investigating.

  1. Perform Analysis, Analyze Patterns, Find Hot Spots
  2. Choose the layer of areas to analyze
  3. Divide the actual counts by expected counts so the statistic uses rates rather than counts. If the data you are using is actual rates rather than counts, you will not need this step
  4. Give the new hot spot layer a meaningful name
  5. Uncheck Use current map extent
  6. Show credits - if this is over 20, you may be analyzing the wrong layer, or you may want to leave Use current map extent checked to reduce the number of areas analyzed
  7. Run Analysis - this may take a few minutes

This tool sometimes freezes and never returns. If your analysis doesn't complete after three minutes:

  1. Check your content page to see if the hot spot layer was created
  2. Reload your map and then add the hot spot layer by hand
Hot Spot Analysis