Hot Spot Analysis

Discussion created by juank52 on Mar 15, 2011
Latest reply on Mar 22, 2011 by lrosenshein-esristaff
Hi Lauren, Im new using the spatial stats tools, so am very excited to use these tools in one f my university projects. So what i did was to run the hot spot analysis for a dataset of districts from my country. I wanted to model the spatial distribution of a socioeconomic variable with the hot spot analysis at district scale.
First, i did some of the common tasks shown in the tutorials that you presented, for example use the "Calculate distance band from neighbor count" tool to ensure that all features had at least one neighbor.
Also, i performed the "Average nearest neighbor" tool to define the increments on the distance bands. Then i graphed the results and chose as the distance band the first peak in tha graph (By the way, it was the only peak, and from there the curve started to go down and down as distances increased). Also I chose as the conceptualization of relationships, the Fixed distance band because the polygon's size varies across the study area.
So im a  little confused, because after obteined the results, i started to run the autocorrelation tool several times, and realized that even at smaller distances than the distance reported by "Calcule distance band from neighbor count''( wich i used as the initial distance) the data still autocorrelated,
im taking distances so small such as 100m and 500m considering that i'm working mainly with large
polygons (Districts polygons for the whole country).
I wonder if am doing something wrong or can i trust  my results (I'll try to send you an image of the hotspot analysis map to give you an idea)

Thank you very much
Any suggestion will be appreciated