Embedded Files
Learning Objectives for Module 6
More often than not, database records contain errors. These errors may result from mis-classifying as a result of ambiguity in the procedure, faulty transcription, inappropriate calculations, or by measurement error. One means of mitigating classification errors is through the use of a confusion matrix. The matrix assumes that these errors have been identified, for example through field verification and results in a percentage of correctly classified values. Since a certain proportion of these are correct by chance, math manipulations help account for the classifications that are correct by chance, like multiplying the row total by the column total and dividing by the grand total and calculating the Kappa Index. Another means of mitigating error, such as with measurements, is to apply statistical approaches, for example root mean square error (RMSE) assessments. The faulty data can subsequently be removed before analysis is performed. Yet another means is via fuzzy approaches, including probability or membership assessments, or simply living with the error.
Problem:
When maps are digitized, the coordinates are in the measurement units used by the digitizing equipment, such as centimeters. These values need to be transformed to standard coordinates. To check the accuracy of the transformation, RMSE analyses are performed.
Analysis Procedure:
After loading the data to an ArcMap document, the Editor and Spatial Adjustment toolbars need to be enabled. In order to adjust the imported data, a control points file needs to be opened and used. In this exercise, a file was provided called gps.txt, and it included GPS control point coordinates of known locations. RMSE was evaluated to identify the best fit between the source and destination control points. With this information, transformation can be performed, followed by definition of a projection.
Results:
Figure 1. Map generated from transformed data
Application and Reflection:
Despite our best efforts to be as conscientious as possible with data collection and data processing, errors or uncertainty are inevitable. Having methods to account for these is a requirement for the production of high quality data that can be used to elucidate the questions we are trying to answer. For the mushroom project (see Our Mushroom Hunting), we might find that the habitat within which the mushrooms are seen is hard to classify. Some days, we might classify the habitat rocky with some detritus, and other days we might classify the habitat covered with detritus with some rocks. In these cases, we might consider the confusion matrix approach, or the fuzzy membership approach.
Page updated
Google Sites
Report abuse