We have developed an on-line and in-field error modeling technique, which is a generalization of the calibration problem, that relies on a small number of inaccurate sensors with known error distributions to develop error models for the deployed in-field sensors. We demonstrate the applicability of our transitive error modeling technique and evaluate its performance in various scenarios by conducting experiments using traces of the light intensity measurements recorded by in-field deployed light sensors. In addition, statistical validation and evaluation methods such as resubsitution are used in order to establish the interval of confidence.