In the data collection stage, we faced the Modifiable Areal Unit Problem when converting vector data to raster. The census data we used was collected through the smallest areal unit possible. Some uncertainty was present when renter data was available only in Census Tract units, when the population data came in Dissemination Area units. The Multi-Criteria Evaluation tool needs the data to be raster format. Therefore, all census data had to be converted from vector to raster format. The boundaries of the DA and CT are very delicate and precise on the map. Even though we have chosen the smallest pixel size within the constraint of 10m, the pixel size doesn’t fully account for the roughness of the conversion. There are boundaries that show straight lines rather than the original boundary shape.

Uncertainty present in the project analysis is the suppression of CT and DA data that occurs when we do a table join between Excel workbooks. There are DAs and CTs where data was suppressed, thus the join by ‘only matching records’ results in the removal of some units, leaving blank locations in the final map.

In our manipulation of the population density data, we found that there are extreme values that occur in areas known to have low density. For example, the data shows higher population density in Bowen Island than Downtown Vancouver. As a result, we have removed Bowen Island from the MCE analysis and left the area as a field of no data.

The last error we encountered was with the large DA of the North Shore Mountains. This areal unit has little to no people residing, as only about 150 people live within the entire DA. This error was corrected by changing the corresponding population density and number of renter data to 0, as well as using a Small membership type.