Since the data for this experiment were generated using a simulator running on a statistical software (S-Plus), the errors in the acquisition process are not an issue as they could be when collecting data in the field (e.g., ecological variables) or from derivated sources. Hence, it is difficult to determine sources of errors. Only two sources can be pointed out: 1) human errors when compiling (not in the computing sense) the results of the slope values, crosstabulation, average distances and rounding in Microsoft Excel, and 2) fundamental problem in the conditional auto-regressive model. The former can be considered marginal since the small amount of data and the few manipulations that have been done could hardly introduce errors, and it is beyond the scope of this project (and the competencies of the author) to assess the conditional auto-regressive model.

However, from this experiment, one can find useful information about the importance of the quality of the data in such an analysis. In that particular case, the sampling design is a predominant factor for data quality. Because it was shown how the window size could have an impact on the edge detection, it was concluded that the scale of the analysis is a major concern. As the scale of analysis is not only determinate by the window size but also by the resolution on the ground during the data collection (or derivation), it is therefore meaningful that one considers a sampling design adapted not only to the observed ecological (or other) variables but also to the research question.