Problems and Discussion
The census data I got were five years old; I could not find the Census 2006 data. Therefore, the analysis was largely suitable for 2001 but it may apply to present situation as well if no large changes occur in the 2006 census data compared to the 2001 data. In addition, I could not find a shapefile in Dissemination Area 2001 that had the same projection system as my other data. I was forced to use the bigger unit, Census Tract instead. The larger census unit could not show more regional variation; thus the resulted preschool locations were big and their size were largely determined by the size of the Census Tract. Besides, there was a ecological fallacy problem in which each census tract represented a homogenous number of attributes, for example, the total number of children under 6. Each census tract assumed there was no variation in number within the tract but in reality population tends to cluster and vary regionally. Moreover, the census data were relevant only to children under 6. Since there was no further data for children under only 5 years, I had to use the census data even though preschools are generally for children age 0 to 5.
Furthermore, I believed the location of hubs was not perfectlly matched to the one in VanMap because I visually transefered the hub locations data from VanMap to ArcGIS by comparing street networks data available in S: Drive and VanMap as reference. I noticed that the street networks in VanMap had a slight different definition of streets than those in S: Drive because some streets were missing in the S:Drive data. Besides, I could not find out the project system used in VanMap and I believed it was different than the one in S: Drive. As a result, the hub locations I created in ArcGIS should be a bit different than the ones in VanMap and error was assumed.
In addition, the weights assigned to each model were arbitrary. The specially heavily weighted factors in the business model were biased. Different people may have different opinion on the choice of weights and factors; therefore, they would have a different result if they choose a different sets of weights and factors.
I should have spent more time collecting data about library locations, other family and child care service centers locations so that the Model 1 result would match more towards the City of Vancouver's Guildlines. However, this process would take a huge amount of time since there are more than two hunderds of such locations.