Discussion
The initial
stages of this study included the creation of spatial dataset with all the
locations of features in Vancouver that would negatively affect Feng Shui. As evident in Figure 1, there are only a few
locations of funeral homes and crematoriums in Vancouver. These locations are
also correlated with areas that are known for low cost real estate and a high
density of retail shops (Kingsway and West Hastings Street). It should be noted
that these are located in areas that are less desirable then most areas in
Vancouver, which could also be correlated to lower sales prices. The location
of care centers however, where more uniformly distributed throughout Vancouver.
These locations though are less obviously negative. While this category includes
hospices where a lot of death occurs, some people may also find living near a
hospital as a positive selling feature.
The data presented
in Table 1 outlines the results of GWR analysis with different input variables.
The variables which were able to best predict the sales price in Vancouver were
‘home size’ and the combination of ‘home size’ with the negative “Feng Shui
Elements’ as described in the Methodology section. An illustration of the geographic
difference in the GWR results between lot size and home size is shown in Figure
2. The greatest difference visible in the greater predictive capability of the
model is in southern Vancouver. This may be attributed to cultural differences of
the people who buy in that area. The availability and size of a yard in this
area may be less important to number of bedrooms, for example. An illustration
of the GWR results between ‘home size’ with and without the negative Feng Shui
elements is shown in Figure 3. Here it is difficult to discern what areas of
Vancouver this model better predicts sale price in. This is not surprising though,
given the small increase in the R² value.
In order to
determine if the areas where this model best predicts sales price is correlated
to areas with a higher population of Chinese residents, the results were layered
over a chloropleth map of ‘Chinese visible Minority’ as determined by the
Canadian 2006 Census (Figure 4). Although no strong correlation are visually apparent, one
may note the region where the model best predicts is also one of a medium
number of Chinese residents (numbers reported by 20% population survey). Given
the hypothesis of this study, one would expect to see the areas with the most
Chinese residents (black chloropleth) with the highest local R² (dark blue
dots). Although the areas with the
highest Chinese population where not included in the study region (as determined
by the sales data availability), not all of the medium number of Chinese areas
where correlated with high R² values.
Given that
past research concluded that the presence of a 4 in an address decreases the
sold price (in specific communities) the output of this model should give an
over estimate; as it does not account for the presence of 4’s. When the average value of the standard residuals
where compared between homes with and without 4’s, a difference was found. Those
homes with 4’s in their address had a positive standard residual (therefore the
observed value was higher than the expected in the model). The homes with no 4’s
however, resulted in an average negative standard residual value (where the
observed value was less than the expected). It may then be concluded then, that
as the expected value generated in this model is less the observed value in
homes with 4’s, it does not over estimate the sold value form homes with fours.
Therefore this study fails to show the presence of a 4 decreases the sold price
of homes in Vancouver. Further analysis of the GWR results was done using the
Students T-Test for significance. This test assesses whether the means of two
groups are statistically different from another. The assumption was made that
the two samples (homes with and without 4’s in their address) have a normal
distribution. The resultant value of 0.673 is not large enough to conclude with
confidence there is a difference in the sample.