The selection of appropriate values or prior distributions for these parameters is very difficult. Although Jeffreys (30) (a more recent discussion is given by (31)) provides a scale for the interpretation of the Bayes Factor, this scale merely provides a descriptive statement for ease of interpretability, rather than facilitating an actual calibration. In addition, despite recent successes in mapping copy number variation in humans (8,32), the high reported false negative rates in these experiments mean that the true length distribution of copy number variants remains unknown and prevents us from adopting semi-Markov type approaches (33,34) which could exploit such knowledge. By adopting the Objective Bayes paradigm, we now have an objective by which to choose appropriate parameter values, in this case, we select parameter values that calibrate our model to given false positive error rates.
