Conducting statistical analysis and being able to generate numbers to define the spatial relationships between features on a map makes the ArcGIS quite invaluable to researchers. The process of this analysis must first be qualified for its limitations. The monocentric city model offers a framework to better understand the actual processes of demand in Boston, but the ring methodology does not factor any spatial-weights in regards to the Boston context. Moreover, the combined analysis comparing spatial location to statistical measures allows one to better assess the discrepancies in access to urban amenities. This project has reaffirmed the monocentric city model in the Boston context and provided insight on the spatial patterning of transit locations around certain types of land use. The following summarizes the main findings and future considerations from this project:

 

1. Regression Model - The regression analysis conducted only offered a very foundational form that would need to be expanded and requires an analytical judgment call on how to define the dependent variable of housing demand in terms of sales quantity or aggregate sale prices. If the model uses the average sale price dependent variable, this specification would lend itself to the incorporation of zonal statistics. For example, the average nearest distance to transport access points from a residential property would influence the sale price. Similarly, this model could include "Count" variables that relate the number of resources in a town to the sale price of a residential property. Refining the model and correctly including variables that will improve fit and significance will be the best in terms of determing spatial significance. Models could be generated for transportation demand as well.

2. Land Use - A residual finding from this research dealt with zoning allocation for land use. The regression model and cluster/outlier analyses both were able to classify locations where housing demand, in terms of quantity, is effected by variables other than the ones included, which were population, area of the census tract, and the average assessment price. Relating housing demand attributes to land use coverage within each analysis ring would show the effects of zoning on housing and transit demands, and as a result, allow for planners to optimally zone for different purposes.

3. Extensions & Forecasting - Most further research would involve refining the layers of analysis that can be done. For example, transportation analysis could be done on a parcel or town basis, while it could also use other non-political boundaries such as the definitions of metropolitan regions and economic analysis zones. Moreover, the creation of spatially-weighted statistics for the land use data layer would help to improve the significance of the findings. Additionally, average nearest distances should be calculated between zone types and these metrics can be compared on a spatial basis. Particularly, the calculation of an average nearest distances of features within the different analysis rings. These would reflect changes in the levels of clustering and dispersion as one gets further away from the central business districts. Also, given that it works with the data, creating random points for each distinct land use polygon would allow one to more easily calculate zonal and near statistics within the software.

4. Network Analyst - ArcGIS contains other tools like the network analyst which would be a perfect fit to this type of research. Using this feature would allow one to input the data from this project and build a network, linking the street grids and highways with the train systems. Moreover, population, demographic, and land use data would be inputted to inform the analysis. This tool also has the ability to create cost matrices for transit systems and can solve weighted route problems, which could model the commuter experience from various points around the Boston area.