To determine the parameters we want (stellar magnitudes and colors) from the observations we have, we fit the data to a model. The model should represent the circumstances of the observations as realistically as possible; otherwise, effects present in the data but not in the model will be distributed among the available adjustable parameters, such as stellar magnitudes, extinction coefficients, and transformation parameters. The result may well be a good fit to the data, with small residuals; but the parameters will be biased -- that is, they will have significant systematic errors. Such errors are called reduction errors by Manfroid and Sterken [16]