When using the COUNT_INCIDENT_WITHIN_FISHNET_POLYGONS Incident Data Aggregation method with the same Input Features and Analysis Field for Optimized Hot Spot Analysis, but with different bounding polygon extents, the results of the analysis are different. This is due to how the bounding polygons define where the incident Input Features occurs. The cluster of incidents are more likely to be identified as hot spots with a larger bounding polygon encompassing the incidents.
Hot Spot Analysis looks at each feature within the context of neighboring features. When the COUNT_INCIDENTS_WITHIN_FISHNET_POLYGONS method is selected and a Bounding Polygons Defining Where Incidents Are Possible parameter is defined, all fishnets intersecting the bounding polygons are retained for analysis. Fishnets with no underlying incidents receive an incident count of zero. So, when working with the same incidents dataset, the bigger the bounding polygon, the more incidents are clustered within the analysis extent. The areas in which the incidents are clustered are more likely to be identified as hot spots.