Problem
Issues
- Methodological
- Error
1 Methodological
The results of this analysis are by no means perfect. I would have
liked to have more data layers available for analysis. One layer
that would have been valueable is a data layer showing traffic flow
volumes for the road networks, preferably having it associated with
a time series. This data was not available. This information would
have boosted the value of my analysis considerably. My analysis
also simplifies the site selection process to include only a few
criteria. In reality there are much more criteria to consider.
A restaurant datatable was not created. It would have been valueable
if I had information on revenues for past quarters and land lease
value for each competitor restaurant so that I could make a datatable
representing this customized data collection. Unfortunetly this
information is not available, and also this would have been tremendous
work gathering this information for all 31 restaurants in my dataset.
Finally, the use of census tracts simplifies the actual distribution
of the demographics. It creates sharp boundaries that divide areas
into crisp demographic regions and that all locations within a census
tract gets represented by a single representative value. I would
have liked to have smaller census regions such as local communities
or neighbourhood blocks where the data will show better representation
of demographic figures. That was one of the reasons why the criteria
of median household income and proportion of families with children
were scrapped early on. The census tracts were too large to show
accurate distribution of these demographic figures.
2 Error
A potential source of error is the burn_census.vct. My digitizng
of the Burnaby census tracts was not accurate due to gaps between
vector polygons. The gaps can clearly be seen when I rasterized
the layer. The raster image had black holes in the display where
there were polygon boundaries that don't match up. This creates
inaccurate regions of value '0' where it should be value '1'. This
layer was not acceptable for creating new data layers using spatial
operations, but it was adequate to display the immediate results
of queries on the datatable in a generated display. Therefore this
image and subsequent images generated form my digitized burn_census.vct
were not used. I avoided this potential error by using the available
layer ct96.vct which was more accurate because it doesn't have any
open gaps.
An error that I know exist but have little control over was the
positional accuracy of my digitized restaurant points in the vector
layer restaurants.vct. For restaurants that were situated in the
malls, it was most problematic since I didn't know how far back
from the road I should digitize them. Usually there is parking space
between the road and the mall. Therefore this problem is reflected
in the histogram for Query 1 which shows a high variance in the
data.
Another source of error has to do with the date of the data. As
mentioned before the demographic data is only accurate up to and
including 1996. Finally, the data layers made available from the
S-Drive are a potential source of error because there is a question
of the accuracy and the age of this data.
|
