Results and Discussion
Ordinary Least Squares Analysis
The OLS models indicate how well the model performed with the red areas signifying portions of under predictions (where the results are higher than the predicted values) and the blue areas represent portions of over prediction (where the results were actually lower than the predicted values by the model). When observing all three categories of crime, the Downtown area and Downtown East side area are both subjected to be under predictions of crime by the model. It seems there are still explanatory variables needed to fully account for the clustering of under predictions within that area. With regards to the OLS for Mischief Crime (Figure 1), it seems this model is less variable with less over/under predictions as well. The adjusted R-squared value for auto theft (Figure3) is only 0.27 which means the explanatory variables only account for 27% of the auto thefts. The OLS for Break and Enters (Figure 2) on the other hand is very variable with a mosaic of different values per DA. This means the model is supposedly performing well as there is no structure at all in the over/under predictions. However, this is not the case as the adjusted R-squared value was found to be 0.16 and the explanatory variables only explained for 16 percent of the break and enters values. The adjusted R-squared of OLS for mischief was 0.21 and accounted for only 21% of its mischief crime again suggesting that the explanatory variables did not encompass all of the factors of crime.
​
​
​
​
​
​
​
​
Figure 1: OLS Mischief
​
Figure 2: OLS Break & Enter
Figure 3: OLS Auto Theft
​
Geographic Weighted Regression Analysis
​
With regards to the GWR of Auto thefts (Figure 4), a clustering of high standard residual is seen surrounding the Downtown and Strathcona/Mt Pleasant maps of Auto theft is seen. The standard residuals indicate areas where the model is most confident in the calculated correlation. For the case of GWR for Auto theft Crime, the correlations between its explanatory factors as a function of auto theft crime are shown adjacent to the GWR map. The high standard residuals value within the Mt Pleasant/Strathcona/Grandview Woodlands suggests that the high correlation between no postsecondary degree, a high adult population and auto theft is existent.
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
Figure 3: GWR Auto Theft
​
​
The GWR for Break and Enter (B&E) crimes (Figure 5) has relatively less clusters of high/low standard residuals as they are spread relatively evenly. However, there are two main areas of higher confidence and standard residuals within Fairview/Mt Pleasant/Strathcona as well as Sunset. When comparing these areas with its correlation counterparts, it suggests that there is a high correlation between adult population and B&E in the Sunset area. Moreover, there is also a high correlation between having no postsecondary education and B&E crimes in the Fairview/Mt Pleasant/Strathcona region. On the contrary, the correlation between housing value and B&E seems to be less prevalent as only sparse locations in east Vancouver were indicated to have higher residuals.
​
​
​
​
​
​
​
​
​
​
​
​
​
Figure 5: GWR Break and Enter
​
​
The GWR for Mischief Crime (Figure 6) clusters its higher residuals within the following locations: Downtown, Mt Pleasant/Fairview, and Sunset. The clusters of higher residuals in Downtown and Mt Pleasant/Fairview suggest that there is a high correlation between unemployment, adult population, and mischief crime especially in Downtown. On the other hand, the Sunset area shows a cluster of medium high residuals which indicate a slight correlation between children population and mischief crime.
​
​
​
​
​
​
​
​
Figure 6: GWR Mischief Crime
Grouping Analysis
​
The Grouping Analysis (Figure 7), through collecting features that exhibit similar characteristics, emphasizes that there are definitely interrelated socioeconomic factors which contribute to the assortment and amount of crime within each spatial area. When comparing this model with the GWR models, the areas of Downtown, Mt Pleasant, Fairview, Sunset, and Strathcona, which fall and occupy the entirety of the green and red areas, could be taken with more confidence.
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
Figure 7: Grouping Analysis
Hotspot Analysis
​
Lastly, the Crime Hotspot (Figure 8) was conducted for referential purposes and displays the highest concentration of crime (Auto theft, Break & Enter, Mischief Crime) within Vancouver. Due to the high density of population within the Downtown core, the concentration and occurrence of crime would also increase leading to the large hotspot of crime indicated on the map. Furthermore, the other smaller hotspot areas of crime are all situated in Fairview, Mt Pleasant, Grandview Woodlands, and Hastings-Sunrise, which are often labeled as districts that are more socioeconomically impoverished within Vancouver.
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
​
Figure 8: Hotspot Analysis
​
​
​
​
