A multiple methodological approach was used to address the research questions proposed in this study.

Data Acquisition & Filtering

After acquiring a wide range of data from a variety of sources the relevant, fields and shapefiles were queried and clipped to the study area and focus. Basic statistics were calculated using Microsoft Excel and ArcMap.

Correlation Analysis & Exploratory Regressions

Correlation analysis was conducted on the two dependant variables, for each of the 11 independent variables. Correlation analysis, completed in Microsoft Excel,  yielded Pearson’s Product-Moment Correlation Co-efficient (PPMC) values for each set of variables. A PPMC is a measure of the linear relationship between a set of variables. A PPMC (r) value can range from +1 to -1. If the yielded value is close to +1 it indicates a strong positive relationship between the variables. Alternatively, if the yielded value is close to -1 it indicates a strong negative relationship between the variables. A yielded value of 0 signifies that there is no relationship between the variables. Correlation analysis was conducted to investigate what variables would be of significance to use in regression analysis. To further the investigation into suitable variables, four exploratory regressions, within ArcMap, were conducted – two for each dependant variable. The first exploratory regression conducted for each dependant variable used all 11 explanatory variables. The report yielded from these first two exploratory regressions contained summaries associated with the inputted explanatory variables. Seeking the set of variables with the highest Adjusted R-Squared value and lowest corrected Akaike Information Criteria value the variables with the strongest relationships were chosen. In both cases, the top five variables were chosen and a second exploratory regression was run on these five variables. Both secondary exploratory regressions, again, yielded the strongest relationships to be notable between all five variables. As such, these five variables were used in regression analysis.

Ordinary Least Squares Regressions & Moran’s I

Ordinary Least Squares regression (OLS) is a global linear regression model, as such it does not account for spatial variation. For the purpose of this research, an OLS regression was conducted for each of the two dependant variables on all five of the explanatory variables. The purpose of conducting the OLS regressions was to determine if important explanatory variables were missing from the analysis. The OLS regression analysis, within ArcMap, yielded residuals which helped to identify areas where values are larger and smaller than the model has estimated

Using the residuals from the OLS regressions, spatial autocorrelation was conducted (Moran’s I). Spatial autocorrelation analysis yields a Moran’s I value which can range from +1 to -1. The closer to 0 the value, the more spatially random it is. Running spatial autocorrelation on the OLS residuals aided in determining if important explanatory variables had been excluded in the OLS, and as such would also be missing when conducting the GWR.

Geographically Weighted Regressions

Geographically Weighted Regression (GWR) is a regression model which accounts for spatial variation. Relationships may differ over space due a number of reasons, including misspecifications og reality, random sampling variation, changing attitudes, and changing preferences. For the purpose of this research GWR analysis, within ArcMap, was used to investigate each phenomena across the city of Toronto.

 Network Analysis

Using the ESRI’s Network Analyst extension, hospital service areas were calculated. This tool required the use of an extensive road network file which was obtained from the Government of Canada for the year 2010. Using this file, a network dataset was created and used in conjunction with the locations of hospitals in Toronto to develop service areas for all twelve hospitals. These service areas depict the distance, by road, in kilometers surrounding the hospitals.