Methodological and Operational Error :
When taking on a project like this the ultimate goal is to produce a meaningful result that can be used to inform people. To ensure this, is to have a clear focus on what you are trying to convey, how you will go about conveying it and ultimately minimizing the amount of error in all of your steps along the way. To understand your project in terms of error allows you to build on the results and make any further analysis more powerful. In general error is either methodological or operational. Operational error is error that arises during the analysis when you are working with the data. Methodological error comes from the misunderstanding, misuse, or complete lack of knowledge prior to and during the analysis.
Operational Error:
In this project operational error may have appeared in a few instances. Although this project was relatively light on geoprocessing, some error may have found its way in via the editing of files in ArcGIS and the subsequent conversion to Idrisi raster format. For example, the clipping of major road and rail layers could potentially lead to edge effect problems. The edge effect is somewhat inherent in this analysis because the only land use data we have was for the Metro Vancouver. This prevented us from creating a buffer into the surrounding areas to ensure the edges truly portrayed the phenomena we are studying. To minimize this I focused the area of output on Vancouver and its immediately surrounding neighbors. My reasoning for this is that approximately 95% of the homeless enumerated in the 2008 count (SPARC, 2008) lived in these areas. The more complex issue of error can be found when looking into the potential methodological errors present in this project.
Literature Based Assumptions:
Although I made every effort to find a data source on which to base my assumptions, it just wasn’t possible in this analysis. I could not find concrete literature supporting my assumptions on the environments in which homeless refuge sites were found. Therefore, much of this project was based on informal conversations with fellow researchers and simply being an observant citizen in Metro Vancouver. This absence of literature was not necessarily a bad thing as it opened my eyes up to the issue surrounding the lack of a true understanding on the spatial factors leading to homeless refuge site selection.
Additional Datasets:
Upon completion of this initial analysis it became apparent that further analyses could be much more robust with the addition of more data sets. Data sets I believe could make this a much more accurate tool (based on literature) include:
- Crime Data (contacts with police, locations and types of crime)
- Health Data (the use of available resources, etc …)
- Weather Data (including rainfall, temperature, etc…)
- BCAA landuse (more detailed than the GVRD_LU layer)
- Census Data
Semantics:
Another issue I found upon reading the literature was the semantic nightmare that is the homelessness issue. The term “hidden homeless” for example was used exhaustively throughout the literature. In some cases it was used to describe homeless people who lived on the street and in others it was used to describe those who were housed, but in very poor conditions (VanWyk & VanWyk, 2005; Fiedler, R., N. Schuurman, and J. Hyndman, 2006). This was just one of several issues of semantics that I uncovered in the literature.
MAUP and EF:
Most geography based analyses should at the very least briefly address these two issues in the analysis of potential error. The Modifiable Area Unit problem, or MAUP, arises when the same data is aggregated at different scales. In this model, however, there was not really any aggregation and thus the effects of the MAUP were minimized. The ecological fallacy occurs when there is an attempt to aggregate data down from a small to a large scale. This too was avoided in this analysis as there was little, or no, aggregation.