Hello, I have been working with census profile data at the Dissemination Area scale from Statistics Canada (significantly smaller than census tracts) using both hot spot analysis and standard choropleth mapping in ArcView. I am comparing Average Income and Average Value of Dwelling in each census between 1996 and 2006 (3 in total), for two different Census Metropolitan Areas in Ontario, Canada. I have mapped the data using the default Jenks Natural Breaks classification(with the default 5 breaks) and have also run the Getis-Ord Gi* Hot Spot Analysis Tool on both variables.
My question is which method is more appropriate to assess spatial patterns in average income and dwelling values over time?
For the Hot Spot Analysis, I have used the Fixed Distance Band Conceptualization (I also used Incremental Spatial Autocorrelation tool to find an appropriate distance band), but the clustering I am most interested in is in the more rural areas of the Census Metropolitan Area, as opposed to the urban core (Dissemination Areas tend to be smaller in the urban core and suburbs and gradually get larger toward the urban fringe and rural areas). It seems that in both choropleth mapping and hot spot analysis results can vary significantly depending on the distance band chosen, or the classification method/number of breaks. I realize there is some subjectivity involved with both methods, but am wondering if anyone can offer thoughts on an appropriate distance band for running the Hot Spot Analysis (or perhaps a different conceptualization of spatial relationships altogether), or whether I should stick to a simple choropleth map for this analysis?
